Move CallBack Server thread creation, initial processing and destruction to RPC Cleanup some RPC code. Remove extraneous fields from nfs41_cb_info and clean up the code. Change KM_SLEEP in mir_nfs41_callback_thread to KM_NOSLEEP. Fix lint warnings
1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright 2008 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 /* 27 * Copyright (c) 1983,1984,1985,1986,1987,1988,1989 AT&T. 28 * All Rights Reserved 29 */ 30 31 #include <sys/param.h> 32 #include <sys/types.h> 33 #include <sys/systm.h> 34 #include <sys/cmn_err.h> 35 #include <sys/vtrace.h> 36 #include <sys/session.h> 37 #include <sys/thread.h> 38 #include <sys/dnlc.h> 39 #include <sys/cred.h> 40 #include <sys/priv.h> 41 #include <sys/list.h> 42 #include <sys/sdt.h> 43 #include <sys/policy.h> 44 45 #include <rpc/types.h> 46 #include <rpc/xdr.h> 47 48 #include <nfs/nfs.h> 49 50 #include <nfs/nfs_clnt.h> 51 52 #include <nfs/nfs4.h> 53 #include <nfs/rnode4.h> 54 #include <nfs/nfs4_clnt.h> 55 #include <nfs/nfs41_sessions.h> 56 #include <nfs/nfs4_clnt_impl.h> 57 58 /* 59 * client side statistics 60 */ 61 static const struct clstat4 clstat4_tmpl = { 62 { "calls", KSTAT_DATA_UINT64 }, 63 { "badcalls", KSTAT_DATA_UINT64 }, 64 { "clgets", KSTAT_DATA_UINT64 }, 65 { "cltoomany", KSTAT_DATA_UINT64 } 66 }; 67 #ifdef DEBUG 68 struct clstat4_debug clstat4_debug = { 69 { "clalloc", KSTAT_DATA_UINT64 }, 70 { "noresponse", KSTAT_DATA_UINT64 }, 71 { "failover", KSTAT_DATA_UINT64 }, 72 { "remap", KSTAT_DATA_UINT64 }, 73 { "nrnode", KSTAT_DATA_UINT64 }, 74 { "access", KSTAT_DATA_UINT64 }, 75 { "dirent", KSTAT_DATA_UINT64 }, 76 { "dirents", KSTAT_DATA_UINT64 }, 77 { "reclaim", KSTAT_DATA_UINT64 }, 78 { "clreclaim", KSTAT_DATA_UINT64 }, 79 { "f_reclaim", KSTAT_DATA_UINT64 }, 80 { "a_reclaim", KSTAT_DATA_UINT64 }, 81 { "r_reclaim", KSTAT_DATA_UINT64 }, 82 { "r_path", KSTAT_DATA_UINT64 } 83 }; 84 #endif 85 86 /* 87 * We keep a global list of per-zone client data, so we can clean up all zones 88 * if we get low on memory. 89 */ 90 static list_t nfs4_clnt_list; 91 static kmutex_t nfs4_clnt_list_lock; 92 93 static struct kmem_cache *chtab4_cache; 94 95 #ifdef DEBUG 96 static int nfs4_rfscall_debug; 97 static int nfs4_try_failover_any; 98 int nfs4_utf8_debug = 0; 99 #endif 100 101 /* 102 * NFSv4 readdir cache implementation 103 */ 104 typedef struct rddir4_cache_impl { 105 rddir4_cache rc; /* readdir cache element */ 106 kmutex_t lock; /* lock protects count */ 107 uint_t count; /* reference count */ 108 avl_node_t tree; /* AVL tree link */ 109 } rddir4_cache_impl; 110 111 static int rddir4_cache_compar(const void *, const void *); 112 static void rddir4_cache_free(rddir4_cache_impl *); 113 static rddir4_cache *rddir4_cache_alloc(int); 114 static void rddir4_cache_hold(rddir4_cache *); 115 static int try_failover(enum clnt_stat); 116 117 static int nfs4_readdir_cache_hits = 0; 118 static int nfs4_readdir_cache_waits = 0; 119 static int nfs4_readdir_cache_misses = 0; 120 121 /* 122 * Shared nfs4 functions 123 */ 124 125 /* 126 * Copy an nfs_fh4. The destination storage (to->nfs_fh4_val) must already 127 * be allocated. 128 */ 129 130 void 131 nfs_fh4_copy(nfs_fh4 *from, nfs_fh4 *to) 132 { 133 to->nfs_fh4_len = from->nfs_fh4_len; 134 bcopy(from->nfs_fh4_val, to->nfs_fh4_val, to->nfs_fh4_len); 135 } 136 137 /* 138 * nfs4cmpfh - compare 2 filehandles. 139 * Returns 0 if the two nfsv4 filehandles are the same, -1 if the first is 140 * "less" than the second, +1 if the first is "greater" than the second. 141 */ 142 143 int 144 nfs4cmpfh(const nfs_fh4 *fh4p1, const nfs_fh4 *fh4p2) 145 { 146 const char *c1, *c2; 147 148 if (fh4p1->nfs_fh4_len < fh4p2->nfs_fh4_len) 149 return (-1); 150 if (fh4p1->nfs_fh4_len > fh4p2->nfs_fh4_len) 151 return (1); 152 for (c1 = fh4p1->nfs_fh4_val, c2 = fh4p2->nfs_fh4_val; 153 c1 < fh4p1->nfs_fh4_val + fh4p1->nfs_fh4_len; 154 c1++, c2++) { 155 if (*c1 < *c2) 156 return (-1); 157 if (*c1 > *c2) 158 return (1); 159 } 160 161 return (0); 162 } 163 164 /* 165 * Compare two v4 filehandles. Return zero if they're the same, non-zero 166 * if they're not. Like nfs4cmpfh(), but different filehandle 167 * representation, and doesn't provide information about greater than or 168 * less than. 169 */ 170 171 int 172 nfs4cmpfhandle(nfs4_fhandle_t *fh1, nfs4_fhandle_t *fh2) 173 { 174 if (fh1->fh_len == fh2->fh_len) 175 return (bcmp(fh1->fh_buf, fh2->fh_buf, fh1->fh_len)); 176 177 return (1); 178 } 179 180 int 181 stateid4_cmp(stateid4 *s1, stateid4 *s2) 182 { 183 if (bcmp(s1, s2, sizeof (stateid4)) == 0) 184 return (1); 185 else 186 return (0); 187 } 188 189 nfsstat4 190 puterrno4(int error) 191 { 192 switch (error) { 193 case 0: 194 return (NFS4_OK); 195 case EPERM: 196 return (NFS4ERR_PERM); 197 case ENOENT: 198 return (NFS4ERR_NOENT); 199 case EINTR: 200 return (NFS4ERR_IO); 201 case EIO: 202 return (NFS4ERR_IO); 203 case ENXIO: 204 return (NFS4ERR_NXIO); 205 case ENOMEM: 206 return (NFS4ERR_RESOURCE); 207 case EACCES: 208 return (NFS4ERR_ACCESS); 209 case EBUSY: 210 return (NFS4ERR_IO); 211 case EEXIST: 212 return (NFS4ERR_EXIST); 213 case EXDEV: 214 return (NFS4ERR_XDEV); 215 case ENODEV: 216 return (NFS4ERR_IO); 217 case ENOTDIR: 218 return (NFS4ERR_NOTDIR); 219 case EISDIR: 220 return (NFS4ERR_ISDIR); 221 case EINVAL: 222 return (NFS4ERR_INVAL); 223 case EMFILE: 224 return (NFS4ERR_RESOURCE); 225 case EFBIG: 226 return (NFS4ERR_FBIG); 227 case ENOSPC: 228 return (NFS4ERR_NOSPC); 229 case EROFS: 230 return (NFS4ERR_ROFS); 231 case EMLINK: 232 return (NFS4ERR_MLINK); 233 case EDEADLK: 234 return (NFS4ERR_DEADLOCK); 235 case ENOLCK: 236 return (NFS4ERR_DENIED); 237 case EREMOTE: 238 return (NFS4ERR_SERVERFAULT); 239 case ENOTSUP: 240 return (NFS4ERR_NOTSUPP); 241 case EDQUOT: 242 return (NFS4ERR_DQUOT); 243 case ENAMETOOLONG: 244 return (NFS4ERR_NAMETOOLONG); 245 case EOVERFLOW: 246 return (NFS4ERR_INVAL); 247 case ENOSYS: 248 return (NFS4ERR_NOTSUPP); 249 case ENOTEMPTY: 250 return (NFS4ERR_NOTEMPTY); 251 case EOPNOTSUPP: 252 return (NFS4ERR_NOTSUPP); 253 case ESTALE: 254 return (NFS4ERR_STALE); 255 case EAGAIN: 256 if (curthread->t_flag & T_WOULDBLOCK) { 257 curthread->t_flag &= ~T_WOULDBLOCK; 258 return (NFS4ERR_DELAY); 259 } 260 return (NFS4ERR_LOCKED); 261 default: 262 return ((enum nfsstat4)error); 263 } 264 } 265 266 int 267 geterrno4(enum nfsstat4 status) 268 { 269 switch (status) { 270 case NFS4_OK: 271 return (0); 272 case NFS4ERR_PERM: 273 return (EPERM); 274 case NFS4ERR_NOENT: 275 return (ENOENT); 276 case NFS4ERR_IO: 277 return (EIO); 278 case NFS4ERR_NXIO: 279 return (ENXIO); 280 case NFS4ERR_ACCESS: 281 return (EACCES); 282 case NFS4ERR_EXIST: 283 return (EEXIST); 284 case NFS4ERR_XDEV: 285 return (EXDEV); 286 case NFS4ERR_NOTDIR: 287 return (ENOTDIR); 288 case NFS4ERR_ISDIR: 289 return (EISDIR); 290 case NFS4ERR_INVAL: 291 return (EINVAL); 292 case NFS4ERR_FBIG: 293 return (EFBIG); 294 case NFS4ERR_NOSPC: 295 return (ENOSPC); 296 case NFS4ERR_ROFS: 297 return (EROFS); 298 case NFS4ERR_MLINK: 299 return (EMLINK); 300 case NFS4ERR_NAMETOOLONG: 301 return (ENAMETOOLONG); 302 case NFS4ERR_NOTEMPTY: 303 return (ENOTEMPTY); 304 case NFS4ERR_DQUOT: 305 return (EDQUOT); 306 case NFS4ERR_STALE: 307 return (ESTALE); 308 case NFS4ERR_BADHANDLE: 309 return (ESTALE); 310 case NFS4ERR_BAD_COOKIE: 311 return (EINVAL); 312 case NFS4ERR_NOTSUPP: 313 return (EOPNOTSUPP); 314 case NFS4ERR_TOOSMALL: 315 return (EINVAL); 316 case NFS4ERR_SERVERFAULT: 317 return (EIO); 318 case NFS4ERR_BADTYPE: 319 return (EINVAL); 320 case NFS4ERR_DELAY: 321 return (ENXIO); 322 case NFS4ERR_SAME: 323 return (EPROTO); 324 case NFS4ERR_DENIED: 325 return (ENOLCK); 326 case NFS4ERR_EXPIRED: 327 return (EPROTO); 328 case NFS4ERR_LOCKED: 329 return (EACCES); 330 case NFS4ERR_GRACE: 331 return (EAGAIN); 332 case NFS4ERR_FHEXPIRED: /* if got here, failed to get a new fh */ 333 return (ESTALE); 334 case NFS4ERR_SHARE_DENIED: 335 return (EACCES); 336 case NFS4ERR_WRONGSEC: 337 return (EPERM); 338 case NFS4ERR_CLID_INUSE: 339 return (EAGAIN); 340 case NFS4ERR_RESOURCE: 341 return (EAGAIN); 342 case NFS4ERR_MOVED: 343 return (EPROTO); 344 case NFS4ERR_NOFILEHANDLE: 345 return (EIO); 346 case NFS4ERR_MINOR_VERS_MISMATCH: 347 return (ENOTSUP); 348 case NFS4ERR_STALE_CLIENTID: 349 return (EIO); 350 case NFS4ERR_STALE_STATEID: 351 return (EIO); 352 case NFS4ERR_OLD_STATEID: 353 return (EIO); 354 case NFS4ERR_BAD_STATEID: 355 return (EIO); 356 case NFS4ERR_BAD_SEQID: 357 return (EIO); 358 case NFS4ERR_NOT_SAME: 359 return (EPROTO); 360 case NFS4ERR_LOCK_RANGE: 361 return (EPROTO); 362 case NFS4ERR_SYMLINK: 363 return (EPROTO); 364 case NFS4ERR_RESTOREFH: 365 return (EPROTO); 366 case NFS4ERR_LEASE_MOVED: 367 return (EPROTO); 368 case NFS4ERR_ATTRNOTSUPP: 369 return (ENOTSUP); 370 case NFS4ERR_NO_GRACE: 371 return (EPROTO); 372 case NFS4ERR_RECLAIM_BAD: 373 return (EPROTO); 374 case NFS4ERR_RECLAIM_CONFLICT: 375 return (EPROTO); 376 case NFS4ERR_BADXDR: 377 return (EINVAL); 378 case NFS4ERR_LOCKS_HELD: 379 return (EIO); 380 case NFS4ERR_OPENMODE: 381 return (EACCES); 382 case NFS4ERR_BADOWNER: 383 /* 384 * Client and server are in different DNS domains 385 * and the NFSMAPID_DOMAIN in /etc/default/nfs 386 * doesn't match. No good answer here. Return 387 * EACCESS, which translates to "permission denied". 388 */ 389 return (EACCES); 390 case NFS4ERR_BADCHAR: 391 return (EINVAL); 392 case NFS4ERR_BADNAME: 393 return (EINVAL); 394 case NFS4ERR_BAD_RANGE: 395 return (EIO); 396 case NFS4ERR_LOCK_NOTSUPP: 397 return (ENOTSUP); 398 case NFS4ERR_OP_ILLEGAL: 399 return (EINVAL); 400 case NFS4ERR_DEADLOCK: 401 return (EDEADLK); 402 case NFS4ERR_FILE_OPEN: 403 return (EACCES); 404 case NFS4ERR_ADMIN_REVOKED: 405 return (EPROTO); 406 case NFS4ERR_CB_PATH_DOWN: 407 return (EPROTO); 408 case NFS4ERR_BADSESSION: 409 return (EIO); 410 default: 411 #ifdef DEBUG 412 zcmn_err(getzoneid(), CE_WARN, "geterrno4: got status %d", 413 status); 414 #endif 415 return ((int)status); 416 } 417 } 418 419 void 420 nfs4_log_badowner(mntinfo4_t *mi, nfs_opnum4 op) 421 { 422 nfs4_server_t *server; 423 424 /* 425 * Return if already printed/queued a msg 426 * for this mount point. 427 */ 428 if (mi->mi_flags & MI4_BADOWNER_DEBUG) 429 return; 430 /* 431 * Happens once per client <-> server pair. 432 */ 433 if (nfs_rw_enter_sig(&mi->mi_recovlock, RW_READER, 434 mi->mi_flags & MI4_INT)) 435 return; 436 437 server = find_nfs4_server(mi); 438 if (server == NULL) { 439 nfs_rw_exit(&mi->mi_recovlock); 440 return; 441 } 442 443 if (!(server->s_flags & N4S_BADOWNER_DEBUG)) { 444 zcmn_err(mi->mi_zone->zone_id, CE_WARN, 445 "!NFSMAPID_DOMAIN does not match" 446 " the server: %s domain.\n" 447 "Please check configuration", 448 mi->mi_curr_serv->sv_hostname); 449 server->s_flags |= N4S_BADOWNER_DEBUG; 450 } 451 mutex_exit(&server->s_lock); 452 nfs4_server_rele(server); 453 nfs_rw_exit(&mi->mi_recovlock); 454 455 /* 456 * Happens once per mntinfo4_t. 457 * This error is deemed as one of the recovery facts "RF_BADOWNER", 458 * queue this in the mesg queue for this mount_info. This message 459 * is not printed, meaning its absent from id_to_dump_solo_fact() 460 * but its there for inspection if the queue is ever dumped/inspected. 461 */ 462 mutex_enter(&mi->mi_lock); 463 if (!(mi->mi_flags & MI4_BADOWNER_DEBUG)) { 464 nfs4_queue_fact(RF_BADOWNER, mi, NFS4ERR_BADOWNER, 0, op, 465 FALSE, NULL, 0, NULL); 466 mi->mi_flags |= MI4_BADOWNER_DEBUG; 467 } 468 mutex_exit(&mi->mi_lock); 469 } 470 471 int 472 nfs4_time_ntov(nfstime4 *ntime, timestruc_t *vatime) 473 { 474 int64_t sec; 475 int32_t nsec; 476 477 /* 478 * Here check that the nfsv4 time is valid for the system. 479 * nfsv4 time value is a signed 64-bit, and the system time 480 * may be either int64_t or int32_t (depends on the kernel), 481 * so if the kernel is 32-bit, the nfsv4 time value may not fit. 482 */ 483 #ifndef _LP64 484 if (! NFS4_TIME_OK(ntime->seconds)) { 485 return (EOVERFLOW); 486 } 487 #endif 488 489 /* Invalid to specify 1 billion (or more) nsecs */ 490 if (ntime->nseconds >= 1000000000) 491 return (EINVAL); 492 493 if (ntime->seconds < 0) { 494 sec = ntime->seconds + 1; 495 nsec = -1000000000 + ntime->nseconds; 496 } else { 497 sec = ntime->seconds; 498 nsec = ntime->nseconds; 499 } 500 501 vatime->tv_sec = sec; 502 vatime->tv_nsec = nsec; 503 504 return (0); 505 } 506 507 int 508 nfs4_time_vton(timestruc_t *vatime, nfstime4 *ntime) 509 { 510 int64_t sec; 511 uint32_t nsec; 512 513 /* 514 * nfsv4 time value is a signed 64-bit, and the system time 515 * may be either int64_t or int32_t (depends on the kernel), 516 * so all system time values will fit. 517 */ 518 if (vatime->tv_nsec >= 0) { 519 sec = vatime->tv_sec; 520 nsec = vatime->tv_nsec; 521 } else { 522 sec = vatime->tv_sec - 1; 523 nsec = 1000000000 + vatime->tv_nsec; 524 } 525 ntime->seconds = sec; 526 ntime->nseconds = nsec; 527 528 return (0); 529 } 530 531 /* 532 * Converts a utf8 string to a valid null terminated filename string. 533 * 534 * XXX - Not actually translating the UTF-8 string as per RFC 2279. 535 * For now, just validate that the UTF-8 string off the wire 536 * does not have characters that will freak out UFS, and leave 537 * it at that. 538 */ 539 char * 540 utf8_to_fn(utf8string *u8s, uint_t *lenp, char *s) 541 { 542 ASSERT(lenp != NULL); 543 544 if (u8s == NULL || u8s->utf8string_len <= 0 || 545 u8s->utf8string_val == NULL) 546 return (NULL); 547 548 /* 549 * Check for obvious illegal filename chars 550 */ 551 if (utf8_strchr(u8s, '/') != NULL) { 552 #ifdef DEBUG 553 if (nfs4_utf8_debug) { 554 char *path; 555 int len = u8s->utf8string_len; 556 557 path = kmem_alloc(len + 1, KM_SLEEP); 558 bcopy(u8s->utf8string_val, path, len); 559 path[len] = '\0'; 560 561 zcmn_err(getzoneid(), CE_WARN, 562 "Invalid UTF-8 filename: %s", path); 563 564 kmem_free(path, len + 1); 565 } 566 #endif 567 return (NULL); 568 } 569 570 return (utf8_to_str(u8s, lenp, s)); 571 } 572 573 /* 574 * Converts a utf8 string to a C string. 575 * kmem_allocs a new string if not supplied 576 */ 577 char * 578 utf8_to_str(utf8string *str, uint_t *lenp, char *s) 579 { 580 char *sp; 581 char *u8p; 582 int len; 583 int i; 584 585 ASSERT(lenp != NULL); 586 587 if (str == NULL) 588 return (NULL); 589 590 u8p = str->utf8string_val; 591 len = str->utf8string_len; 592 if (len <= 0 || u8p == NULL) { 593 if (s) 594 *s = '\0'; 595 return (NULL); 596 } 597 598 sp = s; 599 if (sp == NULL) 600 sp = kmem_alloc(len + 1, KM_SLEEP); 601 602 /* 603 * At least check for embedded nulls 604 */ 605 for (i = 0; i < len; i++) { 606 sp[i] = u8p[i]; 607 if (u8p[i] == '\0') { 608 #ifdef DEBUG 609 zcmn_err(getzoneid(), CE_WARN, 610 "Embedded NULL in UTF-8 string"); 611 #endif 612 if (s == NULL) 613 kmem_free(sp, len + 1); 614 return (NULL); 615 } 616 } 617 sp[len] = '\0'; 618 *lenp = len + 1; 619 620 return (sp); 621 } 622 623 /* 624 * str_to_utf8 - converts a null-terminated C string to a utf8 string 625 */ 626 utf8string * 627 str_to_utf8(char *nm, utf8string *str) 628 { 629 int len; 630 631 if (str == NULL) 632 return (NULL); 633 634 if (nm == NULL || *nm == '\0') { 635 str->utf8string_len = 0; 636 str->utf8string_val = NULL; 637 } 638 639 len = strlen(nm); 640 641 str->utf8string_val = kmem_alloc(len, KM_SLEEP); 642 str->utf8string_len = len; 643 bcopy(nm, str->utf8string_val, len); 644 645 return (str); 646 } 647 648 utf8string * 649 utf8_copy(utf8string *src, utf8string *dest) 650 { 651 if (src == NULL) 652 return (NULL); 653 if (dest == NULL) 654 return (NULL); 655 656 if (src->utf8string_len > 0) { 657 dest->utf8string_val = kmem_alloc(src->utf8string_len, 658 KM_SLEEP); 659 bcopy(src->utf8string_val, dest->utf8string_val, 660 src->utf8string_len); 661 dest->utf8string_len = src->utf8string_len; 662 } else { 663 dest->utf8string_val = NULL; 664 dest->utf8string_len = 0; 665 } 666 667 return (dest); 668 } 669 670 int 671 utf8_compare(const utf8string *a, const utf8string *b) 672 { 673 int mlen, cmp; 674 int alen, blen; 675 char *aval, *bval; 676 677 if ((a == NULL) && (b == NULL)) 678 return (0); 679 else if (a == NULL) 680 return (-1); 681 else if (b == NULL) 682 return (1); 683 684 alen = a->utf8string_len; 685 blen = b->utf8string_len; 686 aval = a->utf8string_val; 687 bval = b->utf8string_val; 688 689 if (((alen == 0) || (aval == NULL)) && 690 ((blen == 0) || (bval == NULL))) 691 return (0); 692 else if ((alen == 0) || (aval == NULL)) 693 return (-1); 694 else if ((blen == 0) || (bval == NULL)) 695 return (1); 696 697 mlen = MIN(alen, blen); 698 cmp = strncmp(aval, bval, mlen); 699 700 if ((cmp == 0) && (alen == blen)) 701 return (0); 702 else if ((cmp == 0) && (alen < blen)) 703 return (-1); 704 else if (cmp == 0) 705 return (1); 706 else if (cmp < 0) 707 return (-1); 708 return (1); 709 } 710 711 /* 712 * utf8_dir_verify - checks that the utf8 string is valid 713 */ 714 int 715 utf8_dir_verify(utf8string *str) 716 { 717 char *nm; 718 int len; 719 720 if (str == NULL) 721 return (0); 722 723 nm = str->utf8string_val; 724 len = str->utf8string_len; 725 if (nm == NULL || len == 0) { 726 return (0); 727 } 728 729 if (len == 1 && nm[0] == '.') 730 return (0); 731 if (len == 2 && nm[0] == '.' && nm[1] == '.') 732 return (0); 733 734 if (utf8_strchr(str, '/') != NULL) 735 return (0); 736 737 if (utf8_strchr(str, '\0') != NULL) 738 return (0); 739 740 return (1); 741 } 742 743 /* 744 * from rpcsec module (common/rpcsec) 745 */ 746 extern int sec_clnt_geth(CLIENT *, struct sec_data *, cred_t *, AUTH **); 747 extern void sec_clnt_freeh(AUTH *); 748 extern void sec_clnt_freeinfo(struct sec_data *); 749 750 /* 751 * authget() gets an auth handle based on the security 752 * information from the servinfo in mountinfo. 753 * The auth handle is stored in ch_client->cl_auth. 754 * 755 * First security flavor of choice is to use sv_secdata 756 * which is initiated by the client. If that fails, get 757 * secinfo from the server and then select one from the 758 * server secinfo list . 759 * 760 * For RPCSEC_GSS flavor, upon success, a secure context is 761 * established between client and server. 762 */ 763 int 764 authget(servinfo4_t *svp, CLIENT *ch_client, cred_t *cr) 765 { 766 int error, i; 767 768 /* 769 * SV4_TRYSECINFO indicates to try the secinfo list from 770 * sv_secinfo until a successful one is reached. Point 771 * sv_currsec to the selected security mechanism for 772 * later sessions. 773 */ 774 (void) nfs_rw_enter_sig(&svp->sv_lock, RW_WRITER, 0); 775 if ((svp->sv_flags & SV4_TRYSECINFO) && svp->sv_secinfo) { 776 for (i = svp->sv_secinfo->index; i < svp->sv_secinfo->count; 777 i++) { 778 if (!(error = sec_clnt_geth(ch_client, 779 &svp->sv_secinfo->sdata[i], 780 cr, &ch_client->cl_auth))) { 781 782 svp->sv_currsec = &svp->sv_secinfo->sdata[i]; 783 svp->sv_secinfo->index = i; 784 /* done */ 785 svp->sv_flags &= ~SV4_TRYSECINFO; 786 break; 787 } 788 789 /* 790 * Allow the caller retry with the security flavor 791 * pointed by svp->sv_secinfo->index when 792 * ETIMEDOUT/ECONNRESET occurs. 793 */ 794 if (error == ETIMEDOUT || error == ECONNRESET) { 795 svp->sv_secinfo->index = i; 796 break; 797 } 798 } 799 } else { 800 /* sv_currsec points to one of the entries in sv_secinfo */ 801 if (svp->sv_currsec) { 802 error = sec_clnt_geth(ch_client, svp->sv_currsec, cr, 803 &ch_client->cl_auth); 804 } else { 805 /* If it's null, use sv_secdata. */ 806 error = sec_clnt_geth(ch_client, svp->sv_secdata, cr, 807 &ch_client->cl_auth); 808 } 809 } 810 nfs_rw_exit(&svp->sv_lock); 811 812 return (error); 813 } 814 815 /* 816 * Common handle get program for NFS, NFS ACL, and NFS AUTH client. 817 */ 818 int 819 clget4(clinfo_t *ci, servinfo4_t *svp, cred_t *cr, CLIENT **newcl, 820 struct chtab **chp, struct nfs4_clnt *nfscl, mntinfo4_t *mi) 821 { 822 struct chhead *ch, *newch; 823 struct chhead **plistp; 824 struct chtab *cp; 825 int error; 826 k_sigset_t smask; 827 828 if (newcl == NULL || chp == NULL || ci == NULL) 829 return (EINVAL); 830 831 *newcl = NULL; 832 *chp = NULL; 833 834 /* 835 * Find an unused handle or create one 836 */ 837 newch = NULL; 838 /* 839 * Update statistics based on minor version number 840 */ 841 nfscl->nfscl_stat[NFS4_MINORVERSION(mi)].clgets.value.ui64++; 842 top: 843 /* 844 * Find the correct entry in the cache to check for free 845 * client handles. The search is based on the RPC program 846 * number, program version number, dev_t for the transport 847 * device, and the protocol family. 848 */ 849 mutex_enter(&nfscl->nfscl_chtable4_lock); 850 plistp = &nfscl->nfscl_chtable4; 851 for (ch = nfscl->nfscl_chtable4; ch != NULL; ch = ch->ch_next) { 852 if (ch->ch_prog == ci->cl_prog && 853 ch->ch_vers == ci->cl_vers && 854 ch->ch_dev == svp->sv_knconf->knc_rdev && 855 (strcmp(ch->ch_protofmly, 856 svp->sv_knconf->knc_protofmly) == 0)) 857 break; 858 plistp = &ch->ch_next; 859 } 860 861 /* 862 * If we didn't find a cache entry for this quadruple, then 863 * create one. If we don't have one already preallocated, 864 * then drop the cache lock, create one, and then start over. 865 * If we did have a preallocated entry, then just add it to 866 * the front of the list. 867 */ 868 if (ch == NULL) { 869 if (newch == NULL) { 870 mutex_exit(&nfscl->nfscl_chtable4_lock); 871 newch = kmem_alloc(sizeof (*newch), KM_SLEEP); 872 newch->ch_timesused = 0; 873 newch->ch_prog = ci->cl_prog; 874 newch->ch_vers = ci->cl_vers; 875 newch->ch_dev = svp->sv_knconf->knc_rdev; 876 newch->ch_protofmly = kmem_alloc( 877 strlen(svp->sv_knconf->knc_protofmly) + 1, 878 KM_SLEEP); 879 (void) strcpy(newch->ch_protofmly, 880 svp->sv_knconf->knc_protofmly); 881 newch->ch_list = NULL; 882 goto top; 883 } 884 ch = newch; 885 newch = NULL; 886 ch->ch_next = nfscl->nfscl_chtable4; 887 nfscl->nfscl_chtable4 = ch; 888 /* 889 * We found a cache entry, but if it isn't on the front of the 890 * list, then move it to the front of the list to try to take 891 * advantage of locality of operations. 892 */ 893 } else if (ch != nfscl->nfscl_chtable4) { 894 *plistp = ch->ch_next; 895 ch->ch_next = nfscl->nfscl_chtable4; 896 nfscl->nfscl_chtable4 = ch; 897 } 898 899 /* 900 * If there was a free client handle cached, then remove it 901 * from the list, init it, and use it. 902 */ 903 if (ch->ch_list != NULL) { 904 cp = ch->ch_list; 905 ch->ch_list = cp->ch_list; 906 mutex_exit(&nfscl->nfscl_chtable4_lock); 907 if (newch != NULL) { 908 kmem_free(newch->ch_protofmly, 909 strlen(newch->ch_protofmly) + 1); 910 kmem_free(newch, sizeof (*newch)); 911 } 912 (void) clnt_tli_kinit(cp->ch_client, svp->sv_knconf, 913 &svp->sv_addr, ci->cl_readsize, ci->cl_retrans, cr); 914 915 /* 916 * Get an auth handle. 917 */ 918 error = authget(svp, cp->ch_client, cr); 919 if (error || cp->ch_client->cl_auth == NULL) { 920 CLNT_DESTROY(cp->ch_client); 921 kmem_cache_free(chtab4_cache, cp); 922 return ((error != 0) ? error : EINTR); 923 } 924 ch->ch_timesused++; 925 *newcl = cp->ch_client; 926 *chp = cp; 927 return (0); 928 } 929 930 /* 931 * There weren't any free client handles which fit, so allocate a 932 * new one and use that. 933 */ 934 #ifdef DEBUG 935 atomic_add_64(&clstat4_debug.clalloc.value.ui64, 1); 936 #endif 937 mutex_exit(&nfscl->nfscl_chtable4_lock); 938 939 nfscl->nfscl_stat[NFS4_MINORVERSION(mi)].cltoomany.value.ui64++; 940 if (newch != NULL) { 941 kmem_free(newch->ch_protofmly, strlen(newch->ch_protofmly) + 1); 942 kmem_free(newch, sizeof (*newch)); 943 } 944 945 cp = kmem_cache_alloc(chtab4_cache, KM_SLEEP); 946 cp->ch_head = ch; 947 948 sigintr(&smask, (int)ci->cl_flags & MI4_INT); 949 error = clnt_tli_kcreate(svp->sv_knconf, &svp->sv_addr, ci->cl_prog, 950 ci->cl_vers, ci->cl_readsize, ci->cl_retrans, cr, &cp->ch_client); 951 sigunintr(&smask); 952 953 if (error != 0) { 954 kmem_cache_free(chtab4_cache, cp); 955 #ifdef DEBUG 956 atomic_add_64(&clstat4_debug.clalloc.value.ui64, -1); 957 #endif 958 /* 959 * Warning is unnecessary if error is EINTR. 960 */ 961 if (error != EINTR) { 962 nfs_cmn_err(error, CE_WARN, 963 "clget: couldn't create handle: %m\n"); 964 } 965 return (error); 966 } 967 (void) CLNT_CONTROL(cp->ch_client, CLSET_PROGRESS, NULL); 968 auth_destroy(cp->ch_client->cl_auth); 969 970 971 972 /* 973 * Get an auth handle. 974 */ 975 error = authget(svp, cp->ch_client, cr); 976 if (error || cp->ch_client->cl_auth == NULL) { 977 CLNT_DESTROY(cp->ch_client); 978 kmem_cache_free(chtab4_cache, cp); 979 #ifdef DEBUG 980 atomic_add_64(&clstat4_debug.clalloc.value.ui64, -1); 981 #endif 982 return ((error != 0) ? error : EINTR); 983 } 984 ch->ch_timesused++; 985 *newcl = cp->ch_client; 986 ASSERT(cp->ch_client->cl_nosignal == FALSE); 987 *chp = cp; 988 return (0); 989 } 990 991 int 992 nfs_clget4(mntinfo4_t *mi, servinfo4_t *svp, cred_t *cr, CLIENT **newcl, 993 struct chtab **chp, struct nfs4_clnt *nfscl) 994 { 995 clinfo_t ci; 996 bool_t is_recov; 997 int firstcall, error = 0; 998 999 /* 1000 * Set read buffer size to rsize 1001 * and add room for RPC headers. 1002 */ 1003 ci.cl_readsize = mi->mi_tsize; 1004 if (ci.cl_readsize != 0) 1005 ci.cl_readsize += (RPC_MAXDATASIZE - NFS_MAXDATA); 1006 1007 /* 1008 * If soft mount and server is down just try once. 1009 * meaning: do not retransmit. 1010 */ 1011 if (!(mi->mi_flags & MI4_HARD) && (mi->mi_flags & MI4_DOWN)) 1012 ci.cl_retrans = 0; 1013 else 1014 ci.cl_retrans = mi->mi_retrans; 1015 1016 ci.cl_prog = mi->mi_prog; 1017 ci.cl_vers = mi->mi_vers; 1018 ci.cl_flags = mi->mi_flags; 1019 1020 /* 1021 * clget4 calls authget() to get an auth handle. For RPCSEC_GSS 1022 * security flavor, the client tries to establish a security context 1023 * by contacting the server. If the connection is timed out or reset, 1024 * e.g. server reboot, we will try again. 1025 */ 1026 1027 /* 1028 * XXXrecovery: We've already captured the nfs4_server_t in 1029 * start_op but we don't (yet) push it down through rfs4call() 1030 * and friends. We need to do that, especially in the case of 1031 * an operation directed to the data server, so that we can 1032 * determine if this thread may be in recovery (non-pNFS, MDS, or DS). 1033 */ 1034 is_recov = (curthread == mi->mi_recovthread); 1035 firstcall = 1; 1036 1037 do { 1038 error = clget4(&ci, svp, cr, newcl, chp, nfscl, mi); 1039 1040 if (error == 0) 1041 break; 1042 1043 /* 1044 * For forced unmount and zone shutdown, bail out but 1045 * let the recovery thread do one more transmission. 1046 */ 1047 if ((FS_OR_ZONE_GONE4(mi->mi_vfsp)) && 1048 (!is_recov || !firstcall)) { 1049 error = EIO; 1050 break; 1051 } 1052 1053 /* do not retry for soft mount */ 1054 if (!(mi->mi_flags & MI4_HARD)) 1055 break; 1056 1057 /* let the caller deal with the failover case */ 1058 if (FAILOVER_MOUNT4(mi)) 1059 break; 1060 1061 firstcall = 0; 1062 1063 } while (error == ETIMEDOUT || error == ECONNRESET); 1064 1065 return (error); 1066 } 1067 1068 void 1069 clfree4(CLIENT *cl, struct chtab *cp, struct nfs4_clnt *nfscl) 1070 { 1071 if (cl->cl_auth != NULL) { 1072 sec_clnt_freeh(cl->cl_auth); 1073 cl->cl_auth = NULL; 1074 } 1075 1076 if (!CLNT_CONTROL(cl, CLSET_TAG_CLEAR, (char *)NULL)) 1077 zcmn_err(getzoneid(), CE_WARN, 1078 "Failed to clear tag on freed client handle"); 1079 1080 if (!(CLNT_CONTROL(cl, CLSET_BACKCHANNEL_CLEAR, NULL))) { 1081 zcmn_err(getzoneid(), CE_WARN, 1082 "Unable to clear backchannel on freed client handle %p", 1083 (void *)cl); 1084 } 1085 1086 /* 1087 * Timestamp this cache entry so that we know when it was last 1088 * used. 1089 */ 1090 cp->ch_freed = gethrestime_sec(); 1091 1092 /* 1093 * Add the free client handle to the front of the list. 1094 * This way, the list will be sorted in youngest to oldest 1095 * order. 1096 */ 1097 mutex_enter(&nfscl->nfscl_chtable4_lock); 1098 cp->ch_list = cp->ch_head->ch_list; 1099 cp->ch_head->ch_list = cp; 1100 mutex_exit(&nfscl->nfscl_chtable4_lock); 1101 } 1102 1103 #define CL_HOLDTIME 60 /* time to hold client handles */ 1104 1105 static void 1106 clreclaim4_zone(struct nfs4_clnt *nfscl, uint_t cl_holdtime) 1107 { 1108 struct chhead *ch; 1109 struct chtab *cp; /* list of objects that can be reclaimed */ 1110 struct chtab *cpe; 1111 struct chtab *cpl; 1112 struct chtab **cpp; 1113 #ifdef DEBUG 1114 int n = 0; 1115 clstat4_debug.clreclaim.value.ui64++; 1116 #endif 1117 1118 /* 1119 * Need to reclaim some memory, so step through the cache 1120 * looking through the lists for entries which can be freed. 1121 */ 1122 cp = NULL; 1123 1124 mutex_enter(&nfscl->nfscl_chtable4_lock); 1125 1126 /* 1127 * Here we step through each non-NULL quadruple and start to 1128 * construct the reclaim list pointed to by cp. Note that 1129 * cp will contain all eligible chtab entries. When this traversal 1130 * completes, chtab entries from the last quadruple will be at the 1131 * front of cp and entries from previously inspected quadruples have 1132 * been appended to the rear of cp. 1133 */ 1134 for (ch = nfscl->nfscl_chtable4; ch != NULL; ch = ch->ch_next) { 1135 if (ch->ch_list == NULL) 1136 continue; 1137 /* 1138 * Search each list for entries older then 1139 * cl_holdtime seconds. The lists are maintained 1140 * in youngest to oldest order so that when the 1141 * first entry is found which is old enough, then 1142 * all of the rest of the entries on the list will 1143 * be old enough as well. 1144 */ 1145 cpl = ch->ch_list; 1146 cpp = &ch->ch_list; 1147 while (cpl != NULL && 1148 cpl->ch_freed + cl_holdtime > gethrestime_sec()) { 1149 cpp = &cpl->ch_list; 1150 cpl = cpl->ch_list; 1151 } 1152 if (cpl != NULL) { 1153 *cpp = NULL; 1154 if (cp != NULL) { 1155 cpe = cpl; 1156 while (cpe->ch_list != NULL) 1157 cpe = cpe->ch_list; 1158 cpe->ch_list = cp; 1159 } 1160 cp = cpl; 1161 } 1162 } 1163 1164 mutex_exit(&nfscl->nfscl_chtable4_lock); 1165 1166 /* 1167 * If cp is empty, then there is nothing to reclaim here. 1168 */ 1169 if (cp == NULL) 1170 return; 1171 1172 /* 1173 * Step through the list of entries to free, destroying each client 1174 * handle and kmem_free'ing the memory for each entry. 1175 */ 1176 while (cp != NULL) { 1177 #ifdef DEBUG 1178 n++; 1179 #endif 1180 CLNT_DESTROY(cp->ch_client); 1181 cpl = cp->ch_list; 1182 kmem_cache_free(chtab4_cache, cp); 1183 cp = cpl; 1184 } 1185 1186 #ifdef DEBUG 1187 /* 1188 * Update clalloc so that nfsstat shows the current number of 1189 * allocated client handles. 1190 */ 1191 atomic_add_64(&clstat4_debug.clalloc.value.ui64, -n); 1192 #endif 1193 } 1194 1195 /* ARGSUSED */ 1196 static void 1197 clreclaim4(void *all) 1198 { 1199 struct nfs4_clnt *nfscl; 1200 1201 /* 1202 * The system is low on memory; go through and try to reclaim some from 1203 * every zone on the system. 1204 */ 1205 mutex_enter(&nfs4_clnt_list_lock); 1206 nfscl = list_head(&nfs4_clnt_list); 1207 for (; nfscl != NULL; nfscl = list_next(&nfs4_clnt_list, nfscl)) 1208 clreclaim4_zone(nfscl, CL_HOLDTIME); 1209 mutex_exit(&nfs4_clnt_list_lock); 1210 } 1211 1212 /* 1213 * Minimum time-out values indexed by call type 1214 * These units are in "eights" of a second to avoid multiplies 1215 */ 1216 static unsigned int minimum_timeo[] = { 1217 6, 7, 10 1218 }; 1219 1220 #define SHORTWAIT (NFS_COTS_TIMEO / 10) 1221 1222 /* 1223 * Back off for retransmission timeout, MAXTIMO is in hz of a sec 1224 */ 1225 #define MAXTIMO (20*hz) 1226 #define backoff(tim) (((tim) < MAXTIMO) ? dobackoff(tim) : (tim)) 1227 #define dobackoff(tim) ((((tim) << 1) > MAXTIMO) ? MAXTIMO : ((tim) << 1)) 1228 1229 static int 1230 nfs4_rfscall(mntinfo4_t *mi, servinfo4_t *svp, 1231 rpcproc_t which, xdrproc_t xdrargs, caddr_t argsp, 1232 xdrproc_t xdrres, caddr_t resp, cred_t *icr, int *doqueue, 1233 enum clnt_stat *rpc_statusp, int flags, struct nfs4_clnt *nfscl) 1234 { 1235 CLIENT *client; 1236 struct chtab *ch; 1237 cred_t *cr = icr; 1238 struct rpc_err rpcerr; 1239 enum clnt_stat status; 1240 int error; 1241 int ctlret; 1242 struct timeval wait; 1243 int timeo; /* in units of hz */ 1244 bool_t tryagain, is_recov; 1245 bool_t cred_cloned = FALSE; 1246 k_sigset_t smask; 1247 #ifdef DEBUG 1248 char *bufp; 1249 #endif 1250 int firstcall; 1251 struct nfs41_cb_info *cbi; 1252 struct nfs4_server *np; 1253 1254 rpcerr.re_status = RPC_SUCCESS; 1255 1256 /* 1257 * If we know that we are rebooting then let's 1258 * not bother with doing any over the wireness. 1259 */ 1260 mutex_enter(&mi->mi_lock); 1261 if (mi->mi_flags & MI4_SHUTDOWN) { 1262 mutex_exit(&mi->mi_lock); 1263 return (EIO); 1264 } 1265 mutex_exit(&mi->mi_lock); 1266 1267 /* For TSOL, use a new cred which has net_mac_aware flag */ 1268 if (!cred_cloned && is_system_labeled()) { 1269 cred_cloned = TRUE; 1270 cr = crdup(icr); 1271 (void) setpflags(NET_MAC_AWARE, 1, cr); 1272 } 1273 1274 /* 1275 * clget() calls clnt_tli_kinit() which clears the xid, so we 1276 * are guaranteed to reprocess the retry as a new request. 1277 */ 1278 if (svp == NULL) 1279 svp = mi->mi_curr_serv; 1280 rpcerr.re_errno = nfs_clget4(mi, svp, cr, &client, &ch, nfscl); 1281 if (rpcerr.re_errno != 0) 1282 return (rpcerr.re_errno); 1283 1284 if (NFS4_MINORVERSION(mi) == 1) { 1285 mutex_enter(&nfs4_server_lst_lock); 1286 np = servinfo4_to_nfs4_server(svp); 1287 mutex_exit(&nfs4_server_lst_lock); 1288 1289 if (np) { 1290 if (np->s_program != 0 && (flags & RFS4CALL_SETCB)) { 1291 cbi = np->zone_globals->nfs4prog2cbinfo 1292 [np->s_program-NFS4_CALLBACK]; 1293 if (cbi != NULL) { 1294 ctlret = CLNT_CONTROL( 1295 client, CLSET_CBSERVER_SETUP, 1296 (char *)&cbi->cb_rpc); 1297 if (ctlret == 0) { 1298 zcmn_err(getzoneid(), CE_WARN, 1299 "Failed to set client" 1300 " handle as callback"); 1301 } 1302 } 1303 1304 if (!np->ssx.bi_rpc) { 1305 ctlret = CLNT_CONTROL(client, 1306 CLSET_BACKCHANNEL, NULL); 1307 if (ctlret == 0) { 1308 zcmn_err(getzoneid(), CE_WARN, 1309 "Failed to set client" 1310 " handle as callback"); 1311 } 1312 } 1313 1314 /* 1315 * In case of non birpc, make sure rpc layer 1316 * reflects the same -- the below call sets 1317 * the RPC flag non birpc. 1318 */ 1319 if (NFS41_CHECK(mi, nfs41_birpc) == FALSE) { 1320 (void) CLNT_CONTROL(client, 1321 CLSET_NON_BIRPC, (char *)NULL); 1322 } 1323 } 1324 1325 if (!CLNT_CONTROL(client, CLSET_TAG, 1326 (char *)(np->ssx.sessionid))) 1327 zcmn_err(getzoneid(), CE_WARN, 1328 "Failed to set tag on client handle"); 1329 1330 mutex_exit(&np->s_lock); 1331 nfs4_server_rele(np); 1332 } 1333 } 1334 1335 timeo = (mi->mi_timeo * hz) / 10; 1336 1337 /* 1338 * If hard mounted fs, retry call forever unless hard error 1339 * occurs. 1340 * 1341 * For forced unmount, let the recovery thread through but return 1342 * an error for all others. This is so that user processes can 1343 * exit quickly. The recovery thread bails out after one 1344 * transmission so that it can tell if it needs to continue. 1345 * 1346 * For zone shutdown, behave as above to encourage quick 1347 * process exit, but also fail quickly when servers have 1348 * timed out before and reduce the timeouts. 1349 */ 1350 1351 /* 1352 * XXXrecovery: We've already captured the nfs4_server_t in 1353 * start_op but we don't (yet) push it down through rfs4call() 1354 * and friends. We need to do that, especially in the case of 1355 * an operation directed to the data server, so that we can 1356 * determine if this thread may be in recovery (non-pNFS, MDS, or DS). 1357 */ 1358 is_recov = (curthread == mi->mi_recovthread); 1359 firstcall = 1; 1360 do { 1361 tryagain = FALSE; 1362 1363 NFS4_DEBUG(nfs4_rfscall_debug, (CE_NOTE, 1364 "nfs4_rfscall: vfs_flag=0x%x, %s", 1365 mi->mi_vfsp->vfs_flag, 1366 is_recov ? "recov thread" : "not recov thread")); 1367 1368 /* 1369 * It's possible while we're retrying the admin 1370 * decided to reboot. 1371 */ 1372 mutex_enter(&mi->mi_lock); 1373 if (mi->mi_flags & MI4_SHUTDOWN) { 1374 mutex_exit(&mi->mi_lock); 1375 clfree4(client, ch, nfscl); 1376 if (cred_cloned) 1377 crfree(cr); 1378 return (EIO); 1379 } 1380 mutex_exit(&mi->mi_lock); 1381 1382 if ((mi->mi_vfsp->vfs_flag & VFS_UNMOUNTED) && 1383 (!is_recov || !firstcall)) { 1384 clfree4(client, ch, nfscl); 1385 if (cred_cloned) 1386 crfree(cr); 1387 return (EIO); 1388 } 1389 1390 if (zone_status_get(curproc->p_zone) >= ZONE_IS_SHUTTING_DOWN) { 1391 mutex_enter(&mi->mi_lock); 1392 if ((mi->mi_flags & MI4_TIMEDOUT) || 1393 !is_recov || !firstcall) { 1394 mutex_exit(&mi->mi_lock); 1395 clfree4(client, ch, nfscl); 1396 if (cred_cloned) 1397 crfree(cr); 1398 return (EIO); 1399 } 1400 mutex_exit(&mi->mi_lock); 1401 timeo = (MIN(mi->mi_timeo, SHORTWAIT) * hz) / 10; 1402 } 1403 1404 firstcall = 0; 1405 TICK_TO_TIMEVAL(timeo, &wait); 1406 1407 /* 1408 * Mask out all signals except SIGHUP, SIGINT, SIGQUIT 1409 * and SIGTERM. (Preserving the existing masks). 1410 * Mask out SIGINT if mount option nointr is specified. 1411 */ 1412 sigintr(&smask, (int)mi->mi_flags & MI4_INT); 1413 if (!(mi->mi_flags & MI4_INT)) 1414 client->cl_nosignal = TRUE; 1415 1416 /* 1417 * If there is a current signal, then don't bother 1418 * even trying to send out the request because we 1419 * won't be able to block waiting for the response. 1420 * Simply assume RPC_INTR and get on with it. 1421 */ 1422 if (ttolwp(curthread) != NULL && ISSIG(curthread, JUSTLOOKING)) 1423 status = RPC_INTR; 1424 else { 1425 status = CLNT_CALL(client, which, xdrargs, argsp, 1426 xdrres, resp, wait); 1427 } 1428 1429 if (!(mi->mi_flags & MI4_INT)) 1430 client->cl_nosignal = FALSE; 1431 /* 1432 * restore original signal mask 1433 */ 1434 sigunintr(&smask); 1435 1436 switch (status) { 1437 case RPC_SUCCESS: 1438 break; 1439 1440 case RPC_INTR: 1441 /* 1442 * There is no way to recover from this error, 1443 * even if mount option nointr is specified. 1444 * SIGKILL, for example, cannot be blocked. 1445 */ 1446 rpcerr.re_status = RPC_INTR; 1447 rpcerr.re_errno = EINTR; 1448 break; 1449 1450 case RPC_CONN_NOT_BOUND: 1451 rpcerr.re_status = status; 1452 rpcerr.re_errno = EIO; 1453 break; 1454 1455 case RPC_UDERROR: 1456 /* 1457 * If the NFS server is local (vold) and 1458 * it goes away then we get RPC_UDERROR. 1459 * This is a retryable error, so we would 1460 * loop, so check to see if the specific 1461 * error was ECONNRESET, indicating that 1462 * target did not exist at all. If so, 1463 * return with RPC_PROGUNAVAIL and 1464 * ECONNRESET to indicate why. 1465 */ 1466 CLNT_GETERR(client, &rpcerr); 1467 if (rpcerr.re_errno == ECONNRESET) { 1468 rpcerr.re_status = RPC_PROGUNAVAIL; 1469 rpcerr.re_errno = ECONNRESET; 1470 break; 1471 } 1472 /*FALLTHROUGH*/ 1473 1474 default: /* probably RPC_TIMEDOUT */ 1475 1476 if (IS_UNRECOVERABLE_RPC(status)) 1477 break; 1478 1479 /* 1480 * increment server not responding count 1481 */ 1482 mutex_enter(&mi->mi_lock); 1483 mi->mi_noresponse++; 1484 mutex_exit(&mi->mi_lock); 1485 #ifdef DEBUG 1486 clstat4_debug.noresponse.value.ui64++; 1487 #endif 1488 /* 1489 * On zone shutdown, mark server dead and move on. 1490 */ 1491 if (zone_status_get(curproc->p_zone) >= 1492 ZONE_IS_SHUTTING_DOWN) { 1493 mutex_enter(&mi->mi_lock); 1494 mi->mi_flags |= MI4_TIMEDOUT; 1495 mutex_exit(&mi->mi_lock); 1496 clfree4(client, ch, nfscl); 1497 if (cred_cloned) 1498 crfree(cr); 1499 return (EIO); 1500 } 1501 1502 /* 1503 * NFS client failover support: 1504 * return and let the caller take care of 1505 * failover. We only return for failover mounts 1506 * because otherwise we want the "not responding" 1507 * message, the timer updates, etc. 1508 */ 1509 if (mi->mi_vers == 4 && FAILOVER_MOUNT4(mi) && 1510 (error = try_failover(status)) != 0) { 1511 clfree4(client, ch, nfscl); 1512 if (cred_cloned) 1513 crfree(cr); 1514 *rpc_statusp = status; 1515 return (error); 1516 } 1517 1518 if (flags & RFSCALL_SOFT) 1519 break; 1520 1521 tryagain = TRUE; 1522 1523 /* 1524 * The call is in progress (over COTS). 1525 * Try the CLNT_CALL again, but don't 1526 * print a noisy error message. 1527 */ 1528 if (status == RPC_INPROGRESS) 1529 break; 1530 1531 timeo = backoff(timeo); 1532 mutex_enter(&mi->mi_lock); 1533 if (!(mi->mi_flags & MI4_PRINTED)) { 1534 mi->mi_flags |= MI4_PRINTED; 1535 mutex_exit(&mi->mi_lock); 1536 nfs4_queue_fact(RF_SRV_NOT_RESPOND, mi, 0, 0, 0, 1537 FALSE, NULL, 0, NULL); 1538 } else 1539 mutex_exit(&mi->mi_lock); 1540 1541 if (*doqueue && nfs_has_ctty()) { 1542 *doqueue = 0; 1543 if (!(mi->mi_flags & MI4_NOPRINT)) 1544 nfs4_queue_fact(RF_SRV_NOT_RESPOND, mi, 1545 0, 0, 0, FALSE, NULL, 0, NULL); 1546 } 1547 } 1548 } while (tryagain); 1549 1550 DTRACE_PROBE2(nfs4__rfscall_debug, enum clnt_stat, status, 1551 int, rpcerr.re_errno); 1552 1553 if (status != RPC_SUCCESS) { 1554 zoneid_t zoneid = mi->mi_zone->zone_id; 1555 1556 /* 1557 * Let soft mounts use the timed out message. 1558 */ 1559 if (status == RPC_INPROGRESS) 1560 status = RPC_TIMEDOUT; 1561 nfscl->nfscl_stat[NFS4_MINORVERSION(mi)].badcalls.value.ui64++; 1562 if (status != RPC_INTR) { 1563 mutex_enter(&mi->mi_lock); 1564 mi->mi_flags |= MI4_DOWN; 1565 mutex_exit(&mi->mi_lock); 1566 CLNT_GETERR(client, &rpcerr); 1567 #ifdef DEBUG 1568 bufp = clnt_sperror(client, svp->sv_hostname); 1569 zprintf(zoneid, "NFS%d %s failed for %s\n", 1570 mi->mi_vers, mi->mi_rfsnames[which], bufp); 1571 if (nfs_has_ctty()) { 1572 if (!(mi->mi_flags & MI4_NOPRINT)) { 1573 uprintf("NFS%d %s failed for %s\n", 1574 mi->mi_vers, mi->mi_rfsnames[which], 1575 bufp); 1576 } 1577 } 1578 kmem_free(bufp, MAXPATHLEN); 1579 #else 1580 zprintf(zoneid, 1581 "NFS %s failed for server %s: error %d (%s)\n", 1582 mi->mi_rfsnames[which], svp->sv_hostname, 1583 status, clnt_sperrno(status)); 1584 if (nfs_has_ctty()) { 1585 if (!(mi->mi_flags & MI4_NOPRINT)) { 1586 uprintf( 1587 "NFS %s failed for server %s: error %d (%s)\n", 1588 mi->mi_rfsnames[which], 1589 svp->sv_hostname, status, 1590 clnt_sperrno(status)); 1591 } 1592 } 1593 #endif 1594 /* 1595 * when CLNT_CALL() fails with RPC_AUTHERROR, 1596 * re_errno is set appropriately depending on 1597 * the authentication error 1598 */ 1599 if (status == RPC_VERSMISMATCH || 1600 status == RPC_PROGVERSMISMATCH) 1601 rpcerr.re_errno = EIO; 1602 } 1603 } else { 1604 /* 1605 * Test the value of mi_down and mi_printed without 1606 * holding the mi_lock mutex. If they are both zero, 1607 * then it is okay to skip the down and printed 1608 * processing. This saves on a mutex_enter and 1609 * mutex_exit pair for a normal, successful RPC. 1610 * This was just complete overhead. 1611 */ 1612 if (mi->mi_flags & (MI4_DOWN | MI4_PRINTED)) { 1613 mutex_enter(&mi->mi_lock); 1614 mi->mi_flags &= ~MI4_DOWN; 1615 if (mi->mi_flags & MI4_PRINTED) { 1616 mi->mi_flags &= ~MI4_PRINTED; 1617 mutex_exit(&mi->mi_lock); 1618 if (!(mi->mi_vfsp->vfs_flag & VFS_UNMOUNTED)) 1619 nfs4_queue_fact(RF_SRV_OK, mi, 0, 0, 1620 0, FALSE, NULL, 0, NULL); 1621 } else 1622 mutex_exit(&mi->mi_lock); 1623 } 1624 1625 if (*doqueue == 0) { 1626 if (!(mi->mi_flags & MI4_NOPRINT) && 1627 !(mi->mi_vfsp->vfs_flag & VFS_UNMOUNTED)) 1628 nfs4_queue_fact(RF_SRV_OK, mi, 0, 0, 0, 1629 FALSE, NULL, 0, NULL); 1630 1631 *doqueue = 1; 1632 } 1633 } 1634 1635 clfree4(client, ch, nfscl); 1636 if (cred_cloned) 1637 crfree(cr); 1638 1639 ASSERT(rpcerr.re_status == RPC_SUCCESS || rpcerr.re_errno != 0); 1640 1641 TRACE_1(TR_FAC_NFS, TR_RFSCALL_END, "nfs4_rfscall_end:errno %d", 1642 rpcerr.re_errno); 1643 1644 *rpc_statusp = status; 1645 return (rpcerr.re_errno); 1646 } 1647 1648 /* 1649 * rfs4call - general wrapper for RPC calls initiated by the client 1650 * KLR-make this a nosequence rfs4call which will not add a sequence op 1651 * XXXrsb - External callers now user rfs4call() with RFS4CALL_NOSEQ. 1652 */ 1653 static void 1654 rfs4call_nosequence(mntinfo4_t *mi, servinfo4_t *svp, COMPOUND4args_clnt *argsp, 1655 COMPOUND4res_clnt *resp, cred_t *cr, int *doqueue, int flags, 1656 nfs4_error_t *ep) 1657 { 1658 int i, error; 1659 enum clnt_stat rpc_status = NFS4_OK; 1660 int num_resops; 1661 struct nfs4_clnt *nfscl; 1662 1663 ASSERT(nfs_zone() == mi->mi_zone); 1664 nfscl = zone_getspecific(nfs4clnt_zone_key, nfs_zone()); 1665 ASSERT(nfscl != NULL); 1666 /* 1667 * Note that the first call will be accounted for the default 1668 * minor version, even if there are no mounts for that minor 1669 * version. The call may result in a minor vesion mismatch and 1670 * subsequent calls will get accounted correctly. It makes sense 1671 * to account the first call for the default minor version, 1672 * because the client thought that this call is for that minor 1673 * version. Same goes for the compound procedure as well. 1674 */ 1675 nfscl->nfscl_stat[NFS4_MINORVERSION(mi)].calls.value.ui64++; 1676 mi->mi_reqs[NFSPROC4_COMPOUND].value.ui64++; 1677 1678 /* XXX - Set up minorversion */ 1679 argsp->minor_vers = NFS4_MINORVERSION(mi); 1680 1681 /* Set up the results struct for XDR usage */ 1682 resp->argsp = argsp; 1683 resp->array = NULL; 1684 resp->status = 0; 1685 resp->decode_len = 0; 1686 1687 error = nfs4_rfscall(mi, svp, NFSPROC4_COMPOUND, 1688 xdr_COMPOUND4args_clnt, (caddr_t)argsp, 1689 xdr_COMPOUND4res_clnt, (caddr_t)resp, cr, 1690 doqueue, &rpc_status, flags, nfscl); 1691 1692 /* 1693 * Map the connection not bound rpc error to nfs 1694 * error. Currently with no connection binding enforcement 1695 * by the client, we won't hit this. With connection binding 1696 * enforcement in the future (with SSV), the below method is 1697 * needed to drive a bind_conn_to_session after a connection 1698 * loss by the client (See section - 2.10.10.1.4 of the draft) 1699 */ 1700 if (error && rpc_status == RPC_CONN_NOT_BOUND) { 1701 ep->error = 0; 1702 ep->rpc_status = 0; 1703 ep->stat = NFS4ERR_CONN_NOT_BOUND_TO_SESSION; 1704 return; 1705 } 1706 1707 /* Return now if it was any other RPC error */ 1708 if (error) { 1709 ep->error = error; 1710 ep->stat = resp->status; 1711 ep->rpc_status = rpc_status; 1712 return; 1713 } 1714 /* 1715 * else we'll count the processed operations. Note that we will 1716 * NOT enter here in case of NFS4ERR_MINOR_VERS_MISMATCH. 1717 */ 1718 num_resops = resp->decode_len; 1719 for (i = 0; i < num_resops; i++) { 1720 /* 1721 * Count the individual operations 1722 * processed by the server. 1723 */ 1724 if (NFS4_MINORVERSION(mi) == NFS4_MINOR_v1) { 1725 if (resp->array[i].resop >= NFSPROC4_NULL && 1726 resp->array[i].resop <= OP_RECLAIM_COMPLETE) { 1727 mi->mi_reqs[resp->array[i].resop].value.ui64++; 1728 } 1729 } else if (NFS4_MINORVERSION(mi) == NFS4_MINOR_v0) { 1730 if (resp->array[i].resop >= NFSPROC4_NULL && 1731 resp->array[i].resop <= OP_RELEASE_LOCKOWNER) { 1732 mi->mi_reqs[resp->array[i].resop].value.ui64++; 1733 } 1734 } 1735 } 1736 1737 ep->error = 0; 1738 ep->stat = resp->status; 1739 ep->rpc_status = rpc_status; 1740 } 1741 1742 void 1743 rfs41_call(mntinfo4_t *mi, servinfo4_t *svp, COMPOUND4args_clnt *argsp, 1744 COMPOUND4res_clnt *resp, cred_t *cr, int *doqueue, int flags, 1745 nfs4_error_t *ep) 1746 { 1747 nfs4_slot_t *slot; 1748 SEQUENCE4res *seqres; 1749 struct nfs4_server *np; 1750 COMPOUND4args_clnt rfs_args, *rfsargp; 1751 COMPOUND4res_clnt rfs_res, *rfsresp; 1752 int add_seq = 0; 1753 1754 /* 1755 * XXXrsb - The following code is likely to change 1756 * For now, we have a pointer from the servinfo4 to the nfs4_server 1757 * If we have a servinfo4 and the pointer is valid, then use it. 1758 * One note, we may have to deal with the "np == NULL" case. 1759 */ 1760 if (svp && svp->sv_ds_n4sp) { 1761 np = svp->sv_ds_n4sp; 1762 nfs4_server_hold(np); 1763 } else { 1764 np = find_nfs4_server(mi); 1765 ASSERT(np != NULL); 1766 mutex_exit(&np->s_lock); 1767 } 1768 1769 1770 /* 1771 * Allocate another args array so we can insert 1772 * a SEQUENCE Op as the first operation, copy already 1773 * built args into it also. 1774 */ 1775 if (argsp->array->argop != OP_SEQUENCE) { 1776 rfs_args.ctag = argsp->ctag; 1777 rfs_args.array_len = argsp->array_len + 1; 1778 rfs_args.array = kmem_zalloc(sizeof (nfs_argop4) * 1779 rfs_args.array_len, KM_SLEEP); 1780 1781 bcopy(argsp->array, rfs_args.array + 1, 1782 sizeof (nfs_argop4) * argsp->array_len); 1783 1784 ASSERT(argsp->array_len >= 1); 1785 rfs_args.array->argop = OP_SEQUENCE; 1786 rfsargp = &rfs_args; 1787 rfsresp = &rfs_res; 1788 add_seq = 1; 1789 } else { 1790 rfsargp = argsp; 1791 rfsresp = resp; 1792 } 1793 1794 /* Set up the sequence OP */ 1795 1796 nfs4sequence_setup(&np->ssx, rfsargp, &slot); 1797 1798 /* 1799 * Send it using rfs4call_nosequence() 1800 * XXXrsb - this will likely be refactored with the rest of 1801 * the rfs4call() family 1802 */ 1803 rfs4call_nosequence(mi, svp, rfsargp, rfsresp, cr, doqueue, flags, ep); 1804 1805 #if 0 1806 zcmn_err(mi->mi_zone->zone_id, CE_WARN, 1807 "Tag: %x SEQUENCE slot: %x seq: %x estatus: %x nstatus: %x", 1808 rfsargp->ctag, 1809 rfsargp->array->nfs_argop4_u.opsequence.sa_slotid, 1810 rfsargp->array->nfs_argop4_u.opsequence.sa_sequenceid, 1811 ep->error, 1812 rfsresp->array != NULL ? 1813 rfsresp->array->nfs_resop4_u.opsequence.status : 0); 1814 1815 if (ep->error || ep->stat || ep->rpc_status) 1816 cmn_err(CE_WARN, "rfs4call failed: %d, %d, %d", 1817 ep->error, ep->stat, ep->rpc_status); 1818 #endif 1819 1820 nfs4sequence_fin(&np->ssx, rfsresp, slot, ep); 1821 1822 /* 1823 * If the OTW call failed completely, or if the 1824 * results array is NULL, just get out 1825 */ 1826 if (ep->error || (ep->stat && rfsresp->array == NULL)) { 1827 1828 if (ep->error == 0) { 1829 ep->error = geterrno4(ep->stat); 1830 } 1831 1832 if (add_seq) 1833 kmem_free(rfs_args.array, 1834 sizeof (nfs_argop4) * rfs_args.array_len); 1835 1836 nfs4_server_rele(np); 1837 return; 1838 } 1839 1840 /* 1841 * Check the results of the sequence op. If it failed and we 1842 * added it for the caller, then we don't have any results 1843 * to return. 1844 */ 1845 seqres = &rfsresp->array->nfs_resop4_u.opsequence; 1846 if (seqres->sr_status != NFS4_OK) { 1847 1848 cmn_err(CE_WARN, "rfs4call: sequence OP failed %d", 1849 seqres->sr_status); 1850 1851 if (add_seq) { 1852 kmem_free(rfs_args.array, 1853 sizeof (nfs_argop4) * rfs_args.array_len); 1854 resp->status = seqres->sr_status; 1855 resp->array_len = resp->decode_len = 0; 1856 resp->array = NULL; 1857 } 1858 /* XXX - xdr_free? free cpy */ 1859 nfs4_server_rele(np); 1860 return; 1861 } 1862 1863 /* 1864 * Update lease time if we have state since SEQUENCE op was successful 1865 */ 1866 mutex_enter(&np->s_lock); 1867 if (np->lease_valid == NFS4_LEASE_VALID && np->state_ref_count) 1868 np->last_renewal_time = gethrestime_sec(); 1869 mutex_exit(&np->s_lock); 1870 1871 /* 1872 * We some results of interest to the, so 1873 * Allocate an additional response array which doesn't have 1874 * SEQUENCE op results, copy results to it if not just a 1875 * SEQUENCE op for lease renewal. 1876 */ 1877 if (add_seq) { 1878 resp->status = rfsresp->status; 1879 resp->array_len = 1880 rfsresp->array_len == 0 ? 0 :rfsresp->array_len - 1; 1881 resp->decode_len = rfsresp->decode_len == 0 ? 0 : 1882 rfsresp->decode_len - 1; 1883 resp->argsp = argsp; 1884 if (resp->array_len > 0) { 1885 ASSERT(rfs_res.array != NULL); 1886 resp->array = 1887 kmem_alloc(sizeof (nfs_resop4) * 1888 resp->array_len, KM_SLEEP); 1889 bcopy(rfsresp->array + 1, resp->array, 1890 sizeof (nfs_resop4) * resp->array_len); 1891 } else { 1892 resp->array = NULL; 1893 } 1894 kmem_free(rfs_args.array, 1895 sizeof (nfs_argop4) * rfs_args.array_len); 1896 kmem_free(rfs_res.array, 1897 sizeof (nfs_resop4) * rfs_res.array_len); 1898 } 1899 nfs4_server_rele(np); 1900 } 1901 1902 void 1903 rfs4call(mntinfo4_t *mi, servinfo4_t *svp, COMPOUND4args_clnt *argsp, 1904 COMPOUND4res_clnt *resp, cred_t *cr, int *doqueue, int flags, 1905 nfs4_error_t *ep) 1906 { 1907 if (NFS4_MINORVERSION(mi) == 0 || (flags & RFS4CALL_NOSEQ)) { 1908 rfs4call_nosequence(mi, svp, argsp, resp, cr, doqueue, 1909 flags, ep); 1910 return; 1911 } 1912 rfs41_call(mi, svp, argsp, resp, cr, doqueue, flags, ep); 1913 } 1914 1915 /* 1916 * nfs4rename_update - updates stored state after a rename. Currently this 1917 * is the path of the object and anything under it, and the filehandle of 1918 * the renamed object. 1919 */ 1920 void 1921 nfs4rename_update(vnode_t *renvp, vnode_t *ndvp, nfs_fh4 *nfh4p, char *nnm) 1922 { 1923 sfh4_update(VTOR4(renvp)->r_fh, nfh4p); 1924 fn_move(VTOSV(renvp)->sv_name, VTOSV(ndvp)->sv_name, nnm); 1925 } 1926 1927 /* 1928 * Routine to look up the filehandle for the given path and rootvp. 1929 * 1930 * Return values: 1931 * - success: returns zero and *statp is set to NFS4_OK, and *fhp is 1932 * updated. 1933 * - error: return value (errno value) and/or *statp is set appropriately. 1934 */ 1935 #define RML_ORDINARY 1 1936 #define RML_NAMED_ATTR 2 1937 #define RML_ATTRDIR 3 1938 1939 static void 1940 remap_lookup(nfs4_fname_t *fname, vnode_t *rootvp, 1941 int filetype, cred_t *cr, 1942 nfs_fh4 *fhp, nfs4_ga_res_t *garp, /* fh, attrs for object */ 1943 nfs_fh4 *pfhp, nfs4_ga_res_t *pgarp, /* fh, attrs for parent */ 1944 nfs4_error_t *ep) 1945 { 1946 COMPOUND4args_clnt args; 1947 COMPOUND4res_clnt res; 1948 nfs_argop4 *argop; 1949 nfs_resop4 *resop; 1950 int num_argops; 1951 lookup4_param_t lookuparg; 1952 nfs_fh4 *tmpfhp; 1953 int doqueue = 1; 1954 char *path; 1955 mntinfo4_t *mi; 1956 1957 ASSERT(fname != NULL); 1958 ASSERT(rootvp->v_type == VDIR); 1959 1960 mi = VTOMI4(rootvp); 1961 path = fn_path(fname); 1962 switch (filetype) { 1963 case RML_NAMED_ATTR: 1964 lookuparg.l4_getattrs = LKP4_LAST_NAMED_ATTR; 1965 args.ctag = TAG_REMAP_LOOKUP_NA; 1966 break; 1967 case RML_ATTRDIR: 1968 lookuparg.l4_getattrs = LKP4_LAST_ATTRDIR; 1969 args.ctag = TAG_REMAP_LOOKUP_AD; 1970 break; 1971 case RML_ORDINARY: 1972 lookuparg.l4_getattrs = LKP4_ALL_ATTRIBUTES; 1973 args.ctag = TAG_REMAP_LOOKUP; 1974 break; 1975 default: 1976 ep->error = EINVAL; 1977 return; 1978 } 1979 lookuparg.argsp = &args; 1980 lookuparg.resp = &res; 1981 lookuparg.header_len = 1; /* Putfh */ 1982 lookuparg.trailer_len = 0; 1983 lookuparg.ga_bits = MI4_DEFAULT_ATTRMAP(mi); 1984 lookuparg.mi = VTOMI4(rootvp); 1985 1986 (void) nfs4lookup_setup(path, &lookuparg, 1); 1987 1988 /* 0: putfh directory */ 1989 argop = args.array; 1990 argop[0].argop = OP_CPUTFH; 1991 argop[0].nfs_argop4_u.opcputfh.sfh = VTOR4(rootvp)->r_fh; 1992 1993 num_argops = args.array_len; 1994 1995 rfs4call(mi, NULL, &args, &res, cr, &doqueue, RFSCALL_SOFT, ep); 1996 1997 if (ep->error || res.status != NFS4_OK) 1998 goto exit; 1999 2000 /* get the object filehandle */ 2001 resop = &res.array[res.array_len - 2]; 2002 if (resop->resop != OP_GETFH) { 2003 nfs4_queue_event(RE_FAIL_REMAP_OP, mi, NULL, 2004 0, NULL, NULL, 0, NULL, 0, TAG_NONE, TAG_NONE, 0, 0); 2005 ep->stat = NFS4ERR_SERVERFAULT; 2006 goto exit; 2007 } 2008 tmpfhp = &resop->nfs_resop4_u.opgetfh.object; 2009 if (tmpfhp->nfs_fh4_len > NFS4_FHSIZE) { 2010 nfs4_queue_event(RE_FAIL_REMAP_LEN, mi, NULL, 2011 tmpfhp->nfs_fh4_len, NULL, NULL, 0, NULL, 0, TAG_NONE, 2012 TAG_NONE, 0, 0); 2013 ep->stat = NFS4ERR_SERVERFAULT; 2014 goto exit; 2015 } 2016 fhp->nfs_fh4_val = kmem_alloc(tmpfhp->nfs_fh4_len, KM_SLEEP); 2017 nfs_fh4_copy(tmpfhp, fhp); 2018 2019 /* get the object attributes */ 2020 resop = &res.array[res.array_len - 1]; 2021 if (garp && resop->resop == OP_GETATTR) 2022 *garp = resop->nfs_resop4_u.opgetattr.ga_res; 2023 2024 /* See if there are enough fields in the response for parent info */ 2025 if ((int)res.array_len - 5 <= 0) 2026 goto exit; 2027 2028 /* get the parent filehandle */ 2029 resop = &res.array[res.array_len - 5]; 2030 if (resop->resop != OP_GETFH) { 2031 nfs4_queue_event(RE_FAIL_REMAP_OP, mi, NULL, 2032 0, NULL, NULL, 0, NULL, 0, TAG_NONE, TAG_NONE, 0, 0); 2033 ep->stat = NFS4ERR_SERVERFAULT; 2034 goto exit; 2035 } 2036 tmpfhp = &resop->nfs_resop4_u.opgetfh.object; 2037 if (tmpfhp->nfs_fh4_len > NFS4_FHSIZE) { 2038 nfs4_queue_event(RE_FAIL_REMAP_LEN, mi, NULL, 2039 tmpfhp->nfs_fh4_len, NULL, NULL, 0, NULL, 0, TAG_NONE, 2040 TAG_NONE, 0, 0); 2041 ep->stat = NFS4ERR_SERVERFAULT; 2042 goto exit; 2043 } 2044 pfhp->nfs_fh4_val = kmem_alloc(tmpfhp->nfs_fh4_len, KM_SLEEP); 2045 nfs_fh4_copy(tmpfhp, pfhp); 2046 2047 /* get the parent attributes */ 2048 resop = &res.array[res.array_len - 4]; 2049 if (pgarp && resop->resop == OP_GETATTR) 2050 *pgarp = resop->nfs_resop4_u.opgetattr.ga_res; 2051 2052 exit: 2053 /* 2054 * It is too hard to remember where all the OP_LOOKUPs are 2055 */ 2056 nfs4args_lookup_free(argop, num_argops); 2057 kmem_free(argop, lookuparg.arglen * sizeof (nfs_argop4)); 2058 2059 if (!ep->error) 2060 (void) xdr_free(xdr_COMPOUND4res_clnt, (caddr_t)&res); 2061 kmem_free(path, strlen(path)+1); 2062 } 2063 2064 /* 2065 * NFS client failover / volatile filehandle support 2066 * 2067 * Recover the filehandle for the given rnode. 2068 * 2069 * Errors are returned via the nfs4_error_t parameter. 2070 */ 2071 2072 void 2073 nfs4_remap_file(mntinfo4_t *mi, vnode_t *vp, int flags, nfs4_error_t *ep) 2074 { 2075 int is_stub; 2076 rnode4_t *rp = VTOR4(vp); 2077 vnode_t *rootvp = NULL; 2078 vnode_t *dvp = NULL; 2079 cred_t *cr, *cred_otw; 2080 nfs4_ga_res_t gar, pgar; 2081 nfs_fh4 newfh = {0, NULL}, newpfh = {0, NULL}; 2082 int filetype = RML_ORDINARY; 2083 nfs4_recov_state_t recov = {NULL, 0, 0}; 2084 int badfhcount = 0; 2085 nfs4_open_stream_t *osp = NULL; 2086 bool_t first_time = TRUE; /* first time getting OTW cred */ 2087 bool_t last_time = FALSE; /* last time getting OTW cred */ 2088 2089 NFS4_DEBUG(nfs4_client_failover_debug, (CE_NOTE, 2090 "nfs4_remap_file: remapping %s", rnode4info(rp))); 2091 ASSERT(nfs4_consistent_type(vp)); 2092 2093 if (vp->v_flag & VROOT) { 2094 nfs4_remap_root(mi, ep, flags); 2095 return; 2096 } 2097 2098 /* 2099 * Given the root fh, use the path stored in 2100 * the rnode to find the fh for the new server. 2101 */ 2102 ep->error = VFS_ROOT(mi->mi_vfsp, &rootvp); 2103 if (ep->error != 0) 2104 return; 2105 2106 cr = curthread->t_cred; 2107 ASSERT(cr != NULL); 2108 get_remap_cred: 2109 /* 2110 * Releases the osp, if it is provided. 2111 * Puts a hold on the cred_otw and the new osp (if found). 2112 */ 2113 cred_otw = nfs4_get_otw_cred_by_osp(rp, cr, &osp, 2114 &first_time, &last_time); 2115 ASSERT(cred_otw != NULL); 2116 2117 if (rp->r_flags & R4ISXATTR) { 2118 filetype = RML_NAMED_ATTR; 2119 (void) vtodv(vp, &dvp, cred_otw, FALSE); 2120 } 2121 2122 if (vp->v_flag & V_XATTRDIR) { 2123 filetype = RML_ATTRDIR; 2124 } 2125 2126 if (filetype == RML_ORDINARY && rootvp->v_type == VREG) { 2127 /* file mount, doesn't need a remap */ 2128 goto done; 2129 } 2130 2131 again: 2132 remap_lookup(rp->r_svnode.sv_name, rootvp, filetype, cred_otw, 2133 &newfh, &gar, &newpfh, &pgar, ep); 2134 2135 NFS4_DEBUG(nfs4_client_failover_debug, (CE_NOTE, 2136 "nfs4_remap_file: remap_lookup returned %d/%d", 2137 ep->error, ep->stat)); 2138 2139 if (last_time == FALSE && ep->error == EACCES) { 2140 crfree(cred_otw); 2141 if (dvp != NULL) 2142 VN_RELE(dvp); 2143 goto get_remap_cred; 2144 } 2145 if (ep->error != 0) 2146 goto done; 2147 2148 switch (ep->stat) { 2149 case NFS4_OK: 2150 badfhcount = 0; 2151 if (recov.rs_flags & NFS4_RS_DELAY_MSG) { 2152 mutex_enter(&rp->r_statelock); 2153 rp->r_delay_interval = 0; 2154 mutex_exit(&rp->r_statelock); 2155 uprintf("NFS File Available..\n"); 2156 } 2157 break; 2158 case NFS4ERR_FHEXPIRED: 2159 case NFS4ERR_BADHANDLE: 2160 /* 2161 * If we ran into filehandle problems, we should try to 2162 * remap the root vnode first and hope life gets better. 2163 * But we need to avoid loops. 2164 */ 2165 if (badfhcount++ > 0) 2166 goto done; 2167 if (newfh.nfs_fh4_len != 0) { 2168 kmem_free(newfh.nfs_fh4_val, newfh.nfs_fh4_len); 2169 newfh.nfs_fh4_len = 0; 2170 } 2171 if (newpfh.nfs_fh4_len != 0) { 2172 kmem_free(newpfh.nfs_fh4_val, newpfh.nfs_fh4_len); 2173 newpfh.nfs_fh4_len = 0; 2174 } 2175 /* relative path - remap rootvp then retry */ 2176 VN_RELE(rootvp); 2177 rootvp = NULL; 2178 nfs4_remap_root(mi, ep, flags); 2179 if (ep->error != 0 || ep->stat != NFS4_OK) 2180 goto done; 2181 ep->error = VFS_ROOT(mi->mi_vfsp, &rootvp); 2182 if (ep->error != 0) 2183 goto done; 2184 goto again; 2185 case NFS4ERR_DELAY: 2186 badfhcount = 0; 2187 nfs4_set_delay_wait(vp); 2188 ep->error = nfs4_wait_for_delay(vp, &recov, 0); 2189 if (ep->error != 0) 2190 goto done; 2191 goto again; 2192 case NFS4ERR_ACCESS: 2193 /* get new cred, try again */ 2194 if (last_time == TRUE) 2195 goto done; 2196 if (dvp != NULL) 2197 VN_RELE(dvp); 2198 crfree(cred_otw); 2199 goto get_remap_cred; 2200 default: 2201 goto done; 2202 } 2203 2204 /* 2205 * Check on the new and old rnodes before updating; 2206 * if the vnode type or size changes, issue a warning 2207 * and mark the file dead. 2208 */ 2209 mutex_enter(&rp->r_statelock); 2210 if (flags & NFS4_REMAP_CKATTRS) { 2211 if (vp->v_type != gar.n4g_va.va_type || 2212 (vp->v_type != VDIR && 2213 rp->r_size != gar.n4g_va.va_size)) { 2214 NFS4_DEBUG(nfs4_client_failover_debug, (CE_NOTE, 2215 "nfs4_remap_file: size %d vs. %d, type %d vs. %d", 2216 (int)rp->r_size, (int)gar.n4g_va.va_size, 2217 vp->v_type, gar.n4g_va.va_type)); 2218 mutex_exit(&rp->r_statelock); 2219 nfs4_queue_event(RE_FILE_DIFF, mi, 2220 rp->r_server->sv_hostname, 0, vp, NULL, 0, NULL, 0, 2221 TAG_NONE, TAG_NONE, 0, 0); 2222 nfs4_fail_recov(vp, NULL, 0, NFS4_OK); 2223 goto done; 2224 } 2225 } 2226 ASSERT(gar.n4g_va.va_type != VNON); 2227 rp->r_server = mi->mi_curr_serv; 2228 2229 /* 2230 * Turn this object into a "stub" object if we 2231 * crossed an underlying server fs boundary. 2232 * 2233 * This stub will be for a mirror-mount. 2234 * 2235 * See comment in r4_do_attrcache() for more details. 2236 */ 2237 is_stub = 0; 2238 if (gar.n4g_fsid_valid) { 2239 (void) nfs_rw_enter_sig(&rp->r_server->sv_lock, RW_READER, 0); 2240 rp->r_srv_fsid = gar.n4g_fsid; 2241 if (!FATTR4_FSID_EQ(&gar.n4g_fsid, &rp->r_server->sv_fsid)) 2242 is_stub = 1; 2243 nfs_rw_exit(&rp->r_server->sv_lock); 2244 #ifdef DEBUG 2245 } else { 2246 NFS4_DEBUG(nfs4_client_failover_debug, (CE_NOTE, 2247 "remap_file: fsid attr not provided by server. rp=%p", 2248 (void *)rp)); 2249 #endif 2250 } 2251 if (is_stub) 2252 r4_stub_mirrormount(rp); 2253 else 2254 r4_stub_none(rp); 2255 mutex_exit(&rp->r_statelock); 2256 nfs4_attrcache_noinval(vp, &gar, gethrtime()); /* force update */ 2257 sfh4_update(rp->r_fh, &newfh); 2258 ASSERT(nfs4_consistent_type(vp)); 2259 2260 /* 2261 * If we got parent info, use it to update the parent 2262 */ 2263 if (newpfh.nfs_fh4_len != 0) { 2264 if (rp->r_svnode.sv_dfh != NULL) 2265 sfh4_update(rp->r_svnode.sv_dfh, &newpfh); 2266 if (dvp != NULL) { 2267 /* force update of attrs */ 2268 nfs4_attrcache_noinval(dvp, &pgar, gethrtime()); 2269 } 2270 } 2271 done: 2272 if (newfh.nfs_fh4_len != 0) 2273 kmem_free(newfh.nfs_fh4_val, newfh.nfs_fh4_len); 2274 if (newpfh.nfs_fh4_len != 0) 2275 kmem_free(newpfh.nfs_fh4_val, newpfh.nfs_fh4_len); 2276 if (cred_otw != NULL) 2277 crfree(cred_otw); 2278 if (rootvp != NULL) 2279 VN_RELE(rootvp); 2280 if (dvp != NULL) 2281 VN_RELE(dvp); 2282 if (osp != NULL) 2283 open_stream_rele(osp, rp); 2284 } 2285 2286 /* 2287 * Client-side failover support: remap the filehandle for vp if it appears 2288 * necessary. errors are returned via the nfs4_error_t parameter; though, 2289 * if there is a problem, we will just try again later. 2290 */ 2291 2292 void 2293 nfs4_check_remap(mntinfo4_t *mi, vnode_t *vp, int flags, nfs4_error_t *ep) 2294 { 2295 if (vp == NULL) 2296 return; 2297 2298 if (!(vp->v_vfsp->vfs_flag & VFS_RDONLY)) 2299 return; 2300 2301 if (VTOR4(vp)->r_server == mi->mi_curr_serv) 2302 return; 2303 2304 nfs4_remap_file(mi, vp, flags, ep); 2305 } 2306 2307 /* 2308 * nfs4_make_dotdot() - find or create a parent vnode of a non-root node. 2309 * 2310 * Our caller has a filehandle for ".." relative to a particular 2311 * directory object. We want to find or create a parent vnode 2312 * with that filehandle and return it. We can of course create 2313 * a vnode from this filehandle, but we need to also make sure 2314 * that if ".." is a regular file (i.e. dvp is a V_XATTRDIR) 2315 * that we have a parent FH for future reopens as well. If 2316 * we have a remap failure, we won't be able to reopen this 2317 * file, but we won't treat that as fatal because a reopen 2318 * is at least unlikely. Someday nfs4_reopen() should look 2319 * for a missing parent FH and try a remap to recover from it. 2320 * 2321 * need_start_op argument indicates whether this function should 2322 * do a start_op before calling remap_lookup(). This should 2323 * be FALSE, if you are the recovery thread or in an op; otherwise, 2324 * set it to TRUE. 2325 */ 2326 int 2327 nfs4_make_dotdot(nfs4_sharedfh_t *fhp, hrtime_t t, vnode_t *dvp, 2328 cred_t *cr, vnode_t **vpp, int need_start_op) 2329 { 2330 mntinfo4_t *mi = VTOMI4(dvp); 2331 nfs4_fname_t *np = NULL, *pnp = NULL; 2332 vnode_t *vp = NULL, *rootvp = NULL; 2333 rnode4_t *rp; 2334 nfs_fh4 newfh = {0, NULL}, newpfh = {0, NULL}; 2335 nfs4_ga_res_t gar, pgar; 2336 vattr_t va, pva; 2337 nfs4_error_t e = { 0, NFS4_OK, RPC_SUCCESS }; 2338 nfs4_sharedfh_t *sfh = NULL, *psfh = NULL; 2339 nfs4_recov_state_t recov_state; 2340 2341 #ifdef DEBUG 2342 /* 2343 * ensure need_start_op is correct 2344 */ 2345 { 2346 int no_need_start_op = (tsd_get(nfs4_tsd_key) || 2347 (curthread == mi->mi_recovthread)); 2348 /* C needs a ^^ operator! */ 2349 ASSERT(((need_start_op) && (!no_need_start_op)) || 2350 ((! need_start_op) && (no_need_start_op))); 2351 } 2352 #endif 2353 ASSERT(VTOMI4(dvp)->mi_zone == nfs_zone()); 2354 2355 NFS4_DEBUG(nfs4_client_shadow_debug, (CE_NOTE, 2356 "nfs4_make_dotdot: called with fhp %p, dvp %s", (void *)fhp, 2357 rnode4info(VTOR4(dvp)))); 2358 2359 /* 2360 * rootvp might be needed eventually. Holding it now will 2361 * ensure that r4find_unlocked() will find it, if ".." is the root. 2362 */ 2363 e.error = VFS_ROOT(mi->mi_vfsp, &rootvp); 2364 if (e.error != 0) 2365 goto out; 2366 rp = r4find_unlocked(fhp, mi->mi_vfsp); 2367 if (rp != NULL) { 2368 *vpp = RTOV4(rp); 2369 VN_RELE(rootvp); 2370 return (0); 2371 } 2372 2373 /* 2374 * Since we don't have the rnode, we have to go over the wire. 2375 * remap_lookup() can get all of the filehandles and attributes 2376 * we need in one operation. 2377 */ 2378 np = fn_parent(VTOSV(dvp)->sv_name); 2379 ASSERT(np != NULL); 2380 2381 recov_state.rs_flags = 0; 2382 recov_state.rs_num_retry_despite_err = 0; 2383 recov_retry: 2384 if (need_start_op) { 2385 e.error = nfs4_start_fop(mi, rootvp, NULL, OH_LOOKUP, 2386 &recov_state, NULL); 2387 if (e.error != 0) { 2388 goto out; 2389 } 2390 } 2391 va.va_type = VNON; 2392 pva.va_type = VNON; 2393 remap_lookup(np, rootvp, RML_ORDINARY, cr, 2394 &newfh, &gar, &newpfh, &pgar, &e); 2395 if (nfs4_needs_recovery(&e, FALSE, mi->mi_vfsp)) { 2396 if (need_start_op) { 2397 bool_t abort; 2398 2399 abort = nfs4_start_recovery(&e, mi, 2400 rootvp, NULL, NULL, NULL, OP_LOOKUP, NULL); 2401 if (abort) { 2402 nfs4_end_fop(mi, rootvp, NULL, OH_LOOKUP, 2403 &recov_state, FALSE); 2404 if (e.error == 0) 2405 e.error = EIO; 2406 goto out; 2407 } 2408 nfs4_end_fop(mi, rootvp, NULL, OH_LOOKUP, 2409 &recov_state, TRUE); 2410 goto recov_retry; 2411 } 2412 if (e.error == 0) 2413 e.error = EIO; 2414 goto out; 2415 } 2416 2417 if (!e.error) { 2418 va = gar.n4g_va; 2419 pva = pgar.n4g_va; 2420 } 2421 2422 if ((e.error != 0) || 2423 (va.va_type != VDIR)) { 2424 if (need_start_op) 2425 nfs4_end_fop(mi, rootvp, NULL, OH_LOOKUP, 2426 &recov_state, FALSE); 2427 if (e.error == 0) 2428 e.error = EIO; 2429 goto out; 2430 } 2431 2432 if (e.stat != NFS4_OK) { 2433 if (need_start_op) 2434 nfs4_end_fop(mi, rootvp, NULL, OH_LOOKUP, 2435 &recov_state, FALSE); 2436 e.error = EIO; 2437 goto out; 2438 } 2439 2440 /* 2441 * It is possible for remap_lookup() to return with no error, 2442 * but without providing the parent filehandle and attrs. 2443 */ 2444 if (pva.va_type != VDIR) { 2445 /* 2446 * Call remap_lookup() again, this time with the 2447 * newpfh and pgar args in the first position. 2448 */ 2449 pnp = fn_parent(np); 2450 if (pnp != NULL) { 2451 remap_lookup(pnp, rootvp, RML_ORDINARY, cr, 2452 &newpfh, &pgar, NULL, NULL, &e); 2453 if (nfs4_needs_recovery(&e, FALSE, 2454 mi->mi_vfsp)) { 2455 if (need_start_op) { 2456 bool_t abort; 2457 2458 abort = nfs4_start_recovery(&e, mi, 2459 rootvp, NULL, NULL, NULL, 2460 OP_LOOKUP, NULL); 2461 if (abort) { 2462 nfs4_end_fop(mi, rootvp, NULL, 2463 OH_LOOKUP, &recov_state, 2464 FALSE); 2465 if (e.error == 0) 2466 e.error = EIO; 2467 goto out; 2468 } 2469 nfs4_end_fop(mi, rootvp, NULL, 2470 OH_LOOKUP, &recov_state, TRUE); 2471 goto recov_retry; 2472 } 2473 if (e.error == 0) 2474 e.error = EIO; 2475 goto out; 2476 } 2477 2478 if (e.stat != NFS4_OK) { 2479 if (need_start_op) 2480 nfs4_end_fop(mi, rootvp, NULL, 2481 OH_LOOKUP, &recov_state, FALSE); 2482 e.error = EIO; 2483 goto out; 2484 } 2485 } 2486 if ((pnp == NULL) || 2487 (e.error != 0) || 2488 (pva.va_type == VNON)) { 2489 if (need_start_op) 2490 nfs4_end_fop(mi, rootvp, NULL, OH_LOOKUP, 2491 &recov_state, FALSE); 2492 if (e.error == 0) 2493 e.error = EIO; 2494 goto out; 2495 } 2496 } 2497 ASSERT(newpfh.nfs_fh4_len != 0); 2498 if (need_start_op) 2499 nfs4_end_fop(mi, rootvp, NULL, OH_LOOKUP, &recov_state, FALSE); 2500 psfh = sfh4_get(&newpfh, mi); 2501 2502 sfh = sfh4_get(&newfh, mi); 2503 vp = makenfs4node_by_fh(sfh, psfh, &np, &gar, mi, cr, t); 2504 2505 out: 2506 if (np != NULL) 2507 fn_rele(&np); 2508 if (pnp != NULL) 2509 fn_rele(&pnp); 2510 if (newfh.nfs_fh4_len != 0) 2511 kmem_free(newfh.nfs_fh4_val, newfh.nfs_fh4_len); 2512 if (newpfh.nfs_fh4_len != 0) 2513 kmem_free(newpfh.nfs_fh4_val, newpfh.nfs_fh4_len); 2514 if (sfh != NULL) 2515 sfh4_rele(&sfh); 2516 if (psfh != NULL) 2517 sfh4_rele(&psfh); 2518 if (rootvp != NULL) 2519 VN_RELE(rootvp); 2520 *vpp = vp; 2521 return (e.error); 2522 } 2523 2524 #ifdef DEBUG 2525 size_t r_path_memuse = 0; 2526 #endif 2527 2528 /* 2529 * NFS client failover support 2530 * 2531 * sv4_free() frees the malloc'd portion of a "servinfo_t". 2532 */ 2533 void 2534 sv4_free(servinfo4_t *svp) 2535 { 2536 servinfo4_t *next; 2537 struct knetconfig *knconf; 2538 2539 while (svp != NULL) { 2540 next = svp->sv_next; 2541 if (svp->sv_dhsec) 2542 sec_clnt_freeinfo(svp->sv_dhsec); 2543 if (svp->sv_secdata) 2544 sec_clnt_freeinfo(svp->sv_secdata); 2545 if (svp->sv_save_secinfo && 2546 svp->sv_save_secinfo != svp->sv_secinfo) 2547 secinfo_free(svp->sv_save_secinfo); 2548 if (svp->sv_secinfo) 2549 secinfo_free(svp->sv_secinfo); 2550 if (svp->sv_hostname && svp->sv_hostnamelen > 0) 2551 kmem_free(svp->sv_hostname, svp->sv_hostnamelen); 2552 knconf = svp->sv_knconf; 2553 if (knconf != NULL) { 2554 if (knconf->knc_protofmly != NULL) 2555 kmem_free(knconf->knc_protofmly, KNC_STRSIZE); 2556 if (knconf->knc_proto != NULL) 2557 kmem_free(knconf->knc_proto, KNC_STRSIZE); 2558 kmem_free(knconf, sizeof (*knconf)); 2559 } 2560 knconf = svp->sv_origknconf; 2561 if (knconf != NULL) { 2562 if (knconf->knc_protofmly != NULL) 2563 kmem_free(knconf->knc_protofmly, KNC_STRSIZE); 2564 if (knconf->knc_proto != NULL) 2565 kmem_free(knconf->knc_proto, KNC_STRSIZE); 2566 kmem_free(knconf, sizeof (*knconf)); 2567 } 2568 if (svp->sv_addr.buf != NULL && svp->sv_addr.maxlen != 0) 2569 kmem_free(svp->sv_addr.buf, svp->sv_addr.maxlen); 2570 if (svp->sv_path != NULL) { 2571 kmem_free(svp->sv_path, svp->sv_pathlen); 2572 } 2573 nfs_rw_destroy(&svp->sv_lock); 2574 2575 /* 2576 * If we have an nfs4_server from a pnfs data server... 2577 * XXXrsb This may go away or change 2578 */ 2579 if (svp->sv_ds_n4sp) 2580 nfs4_server_rele(svp->sv_ds_n4sp); 2581 2582 kmem_free(svp, sizeof (*svp)); 2583 svp = next; 2584 } 2585 } 2586 2587 void 2588 nfs4_printfhandle(nfs4_fhandle_t *fhp) 2589 { 2590 int *ip; 2591 char *buf; 2592 size_t bufsize; 2593 char *cp; 2594 2595 /* 2596 * 13 == "(file handle:" 2597 * maximum of NFS_FHANDLE / sizeof (*ip) elements in fh_buf times 2598 * 1 == ' ' 2599 * 8 == maximum strlen of "%x" 2600 * 3 == ")\n\0" 2601 */ 2602 bufsize = 13 + ((NFS_FHANDLE_LEN / sizeof (*ip)) * (1 + 8)) + 3; 2603 buf = kmem_alloc(bufsize, KM_NOSLEEP); 2604 if (buf == NULL) 2605 return; 2606 2607 cp = buf; 2608 (void) strcpy(cp, "(file handle:"); 2609 while (*cp != '\0') 2610 cp++; 2611 for (ip = (int *)fhp->fh_buf; 2612 ip < (int *)&fhp->fh_buf[fhp->fh_len]; 2613 ip++) { 2614 (void) sprintf(cp, " %x", *ip); 2615 while (*cp != '\0') 2616 cp++; 2617 } 2618 (void) strcpy(cp, ")\n"); 2619 2620 zcmn_err(getzoneid(), CE_CONT, "%s", buf); 2621 2622 kmem_free(buf, bufsize); 2623 } 2624 2625 /* 2626 * The NFSv4 readdir cache subsystem. 2627 * 2628 * We provide a set of interfaces to allow the rest of the system to utilize 2629 * a caching mechanism while encapsulating the details of the actual 2630 * implementation. This should allow for better maintainability and 2631 * extensibility by consolidating the implementation details in one location. 2632 */ 2633 2634 /* 2635 * Comparator used by AVL routines. 2636 */ 2637 static int 2638 rddir4_cache_compar(const void *x, const void *y) 2639 { 2640 rddir4_cache_impl *ai = (rddir4_cache_impl *)x; 2641 rddir4_cache_impl *bi = (rddir4_cache_impl *)y; 2642 rddir4_cache *a = &ai->rc; 2643 rddir4_cache *b = &bi->rc; 2644 2645 if (a->nfs4_cookie == b->nfs4_cookie) { 2646 if (a->buflen == b->buflen) 2647 return (0); 2648 if (a->buflen < b->buflen) 2649 return (-1); 2650 return (1); 2651 } 2652 2653 if (a->nfs4_cookie < b->nfs4_cookie) 2654 return (-1); 2655 2656 return (1); 2657 } 2658 2659 /* 2660 * Allocate an opaque handle for the readdir cache. 2661 */ 2662 void 2663 rddir4_cache_create(rnode4_t *rp) 2664 { 2665 ASSERT(rp->r_dir == NULL); 2666 2667 rp->r_dir = kmem_alloc(sizeof (avl_tree_t), KM_SLEEP); 2668 2669 avl_create(rp->r_dir, rddir4_cache_compar, sizeof (rddir4_cache_impl), 2670 offsetof(rddir4_cache_impl, tree)); 2671 } 2672 2673 /* 2674 * Purge the cache of all cached readdir responses. 2675 */ 2676 void 2677 rddir4_cache_purge(rnode4_t *rp) 2678 { 2679 rddir4_cache_impl *rdip; 2680 rddir4_cache_impl *nrdip; 2681 2682 ASSERT(MUTEX_HELD(&rp->r_statelock)); 2683 2684 if (rp->r_dir == NULL) 2685 return; 2686 2687 rdip = avl_first(rp->r_dir); 2688 2689 while (rdip != NULL) { 2690 nrdip = AVL_NEXT(rp->r_dir, rdip); 2691 avl_remove(rp->r_dir, rdip); 2692 rdip->rc.flags &= ~RDDIRCACHED; 2693 rddir4_cache_rele(rp, &rdip->rc); 2694 rdip = nrdip; 2695 } 2696 ASSERT(avl_numnodes(rp->r_dir) == 0); 2697 } 2698 2699 /* 2700 * Destroy the readdir cache. 2701 */ 2702 void 2703 rddir4_cache_destroy(rnode4_t *rp) 2704 { 2705 ASSERT(MUTEX_HELD(&rp->r_statelock)); 2706 if (rp->r_dir == NULL) 2707 return; 2708 2709 rddir4_cache_purge(rp); 2710 avl_destroy(rp->r_dir); 2711 kmem_free(rp->r_dir, sizeof (avl_tree_t)); 2712 rp->r_dir = NULL; 2713 } 2714 2715 /* 2716 * Locate a readdir response from the readdir cache. 2717 * 2718 * Return values: 2719 * 2720 * NULL - If there is an unrecoverable situation like the operation may have 2721 * been interrupted. 2722 * 2723 * rddir4_cache * - A pointer to a rddir4_cache is returned to the caller. 2724 * The flags are set approprately, such that the caller knows 2725 * what state the entry is in. 2726 */ 2727 rddir4_cache * 2728 rddir4_cache_lookup(rnode4_t *rp, offset_t cookie, int count) 2729 { 2730 rddir4_cache_impl *rdip = NULL; 2731 rddir4_cache_impl srdip; 2732 rddir4_cache *srdc; 2733 rddir4_cache *rdc = NULL; 2734 rddir4_cache *nrdc = NULL; 2735 avl_index_t where; 2736 2737 top: 2738 ASSERT(nfs_rw_lock_held(&rp->r_rwlock, RW_READER)); 2739 ASSERT(MUTEX_HELD(&rp->r_statelock)); 2740 /* 2741 * Check to see if the readdir cache has been disabled. If so, then 2742 * simply allocate an rddir4_cache entry and return it, since caching 2743 * operations do not apply. 2744 */ 2745 if (rp->r_dir == NULL) { 2746 if (nrdc == NULL) { 2747 /* 2748 * Drop the lock because we are doing a sleeping 2749 * allocation. 2750 */ 2751 mutex_exit(&rp->r_statelock); 2752 rdc = rddir4_cache_alloc(KM_SLEEP); 2753 rdc->nfs4_cookie = cookie; 2754 rdc->buflen = count; 2755 mutex_enter(&rp->r_statelock); 2756 return (rdc); 2757 } 2758 return (nrdc); 2759 } 2760 2761 srdc = &srdip.rc; 2762 srdc->nfs4_cookie = cookie; 2763 srdc->buflen = count; 2764 2765 rdip = avl_find(rp->r_dir, &srdip, &where); 2766 2767 /* 2768 * If we didn't find an entry then create one and insert it 2769 * into the cache. 2770 */ 2771 if (rdip == NULL) { 2772 /* 2773 * Check for the case where we have made a second pass through 2774 * the cache due to a lockless allocation. If we find that no 2775 * thread has already inserted this entry, do the insert now 2776 * and return. 2777 */ 2778 if (nrdc != NULL) { 2779 avl_insert(rp->r_dir, nrdc->data, where); 2780 nrdc->flags |= RDDIRCACHED; 2781 rddir4_cache_hold(nrdc); 2782 return (nrdc); 2783 } 2784 2785 #ifdef DEBUG 2786 nfs4_readdir_cache_misses++; 2787 #endif 2788 /* 2789 * First, try to allocate an entry without sleeping. If that 2790 * fails then drop the lock and do a sleeping allocation. 2791 */ 2792 nrdc = rddir4_cache_alloc(KM_NOSLEEP); 2793 if (nrdc != NULL) { 2794 nrdc->nfs4_cookie = cookie; 2795 nrdc->buflen = count; 2796 avl_insert(rp->r_dir, nrdc->data, where); 2797 nrdc->flags |= RDDIRCACHED; 2798 rddir4_cache_hold(nrdc); 2799 return (nrdc); 2800 } 2801 2802 /* 2803 * Drop the lock and do a sleeping allocation. We incur 2804 * additional overhead by having to search the cache again, 2805 * but this case should be rare. 2806 */ 2807 mutex_exit(&rp->r_statelock); 2808 nrdc = rddir4_cache_alloc(KM_SLEEP); 2809 nrdc->nfs4_cookie = cookie; 2810 nrdc->buflen = count; 2811 mutex_enter(&rp->r_statelock); 2812 /* 2813 * We need to take another pass through the cache 2814 * since we dropped our lock to perform the alloc. 2815 * Another thread may have come by and inserted the 2816 * entry we are interested in. 2817 */ 2818 goto top; 2819 } 2820 2821 /* 2822 * Check to see if we need to free our entry. This can happen if 2823 * another thread came along beat us to the insert. We can 2824 * safely call rddir4_cache_free directly because no other thread 2825 * would have a reference to this entry. 2826 */ 2827 if (nrdc != NULL) 2828 rddir4_cache_free((rddir4_cache_impl *)nrdc->data); 2829 2830 #ifdef DEBUG 2831 nfs4_readdir_cache_hits++; 2832 #endif 2833 /* 2834 * Found something. Make sure it's ready to return. 2835 */ 2836 rdc = &rdip->rc; 2837 rddir4_cache_hold(rdc); 2838 /* 2839 * If the cache entry is in the process of being filled in, wait 2840 * until this completes. The RDDIRWAIT bit is set to indicate that 2841 * someone is waiting and when the thread currently filling the entry 2842 * is done, it should do a cv_broadcast to wakeup all of the threads 2843 * waiting for it to finish. If the thread wakes up to find that 2844 * someone new is now trying to complete the the entry, go back 2845 * to sleep. 2846 */ 2847 while (rdc->flags & RDDIR) { 2848 /* 2849 * The entry is not complete. 2850 */ 2851 nfs_rw_exit(&rp->r_rwlock); 2852 rdc->flags |= RDDIRWAIT; 2853 #ifdef DEBUG 2854 nfs4_readdir_cache_waits++; 2855 #endif 2856 while (rdc->flags & RDDIRWAIT) { 2857 if (!cv_wait_sig(&rdc->cv, &rp->r_statelock)) { 2858 /* 2859 * We got interrupted, probably the user 2860 * typed ^C or an alarm fired. We free the 2861 * new entry if we allocated one. 2862 */ 2863 rddir4_cache_rele(rp, rdc); 2864 mutex_exit(&rp->r_statelock); 2865 (void) nfs_rw_enter_sig(&rp->r_rwlock, 2866 RW_READER, FALSE); 2867 mutex_enter(&rp->r_statelock); 2868 return (NULL); 2869 } 2870 } 2871 mutex_exit(&rp->r_statelock); 2872 (void) nfs_rw_enter_sig(&rp->r_rwlock, 2873 RW_READER, FALSE); 2874 mutex_enter(&rp->r_statelock); 2875 } 2876 2877 /* 2878 * The entry we were waiting on may have been purged from 2879 * the cache and should no longer be used, release it and 2880 * start over. 2881 */ 2882 if (!(rdc->flags & RDDIRCACHED)) { 2883 rddir4_cache_rele(rp, rdc); 2884 goto top; 2885 } 2886 2887 /* 2888 * The entry is completed. Return it. 2889 */ 2890 return (rdc); 2891 } 2892 2893 /* 2894 * Allocate a cache element and return it. Can return NULL if memory is 2895 * low. 2896 */ 2897 static rddir4_cache * 2898 rddir4_cache_alloc(int flags) 2899 { 2900 rddir4_cache_impl *rdip = NULL; 2901 rddir4_cache *rc = NULL; 2902 2903 rdip = kmem_alloc(sizeof (rddir4_cache_impl), flags); 2904 2905 if (rdip != NULL) { 2906 rc = &rdip->rc; 2907 rc->data = (void *)rdip; 2908 rc->nfs4_cookie = 0; 2909 rc->nfs4_ncookie = 0; 2910 rc->entries = NULL; 2911 rc->eof = 0; 2912 rc->entlen = 0; 2913 rc->buflen = 0; 2914 rc->actlen = 0; 2915 /* 2916 * A readdir is required so set the flag. 2917 */ 2918 rc->flags = RDDIRREQ; 2919 cv_init(&rc->cv, NULL, CV_DEFAULT, NULL); 2920 rc->error = 0; 2921 mutex_init(&rdip->lock, NULL, MUTEX_DEFAULT, NULL); 2922 rdip->count = 1; 2923 #ifdef DEBUG 2924 atomic_add_64(&clstat4_debug.dirent.value.ui64, 1); 2925 #endif 2926 } 2927 return (rc); 2928 } 2929 2930 /* 2931 * Increment the reference count to this cache element. 2932 */ 2933 static void 2934 rddir4_cache_hold(rddir4_cache *rc) 2935 { 2936 rddir4_cache_impl *rdip = (rddir4_cache_impl *)rc->data; 2937 2938 mutex_enter(&rdip->lock); 2939 rdip->count++; 2940 mutex_exit(&rdip->lock); 2941 } 2942 2943 /* 2944 * Release a reference to this cache element. If the count is zero then 2945 * free the element. 2946 */ 2947 void 2948 rddir4_cache_rele(rnode4_t *rp, rddir4_cache *rdc) 2949 { 2950 rddir4_cache_impl *rdip = (rddir4_cache_impl *)rdc->data; 2951 2952 ASSERT(MUTEX_HELD(&rp->r_statelock)); 2953 2954 /* 2955 * Check to see if we have any waiters. If so, we can wake them 2956 * so that they can proceed. 2957 */ 2958 if (rdc->flags & RDDIRWAIT) { 2959 rdc->flags &= ~RDDIRWAIT; 2960 cv_broadcast(&rdc->cv); 2961 } 2962 2963 mutex_enter(&rdip->lock); 2964 ASSERT(rdip->count > 0); 2965 if (--rdip->count == 0) { 2966 mutex_exit(&rdip->lock); 2967 rddir4_cache_free(rdip); 2968 } else 2969 mutex_exit(&rdip->lock); 2970 } 2971 2972 /* 2973 * Free a cache element. 2974 */ 2975 static void 2976 rddir4_cache_free(rddir4_cache_impl *rdip) 2977 { 2978 rddir4_cache *rc = &rdip->rc; 2979 2980 #ifdef DEBUG 2981 atomic_add_64(&clstat4_debug.dirent.value.ui64, -1); 2982 #endif 2983 if (rc->entries != NULL) 2984 kmem_free(rc->entries, rc->buflen); 2985 cv_destroy(&rc->cv); 2986 mutex_destroy(&rdip->lock); 2987 kmem_free(rdip, sizeof (*rdip)); 2988 } 2989 2990 /* 2991 * Snapshot callback for nfs:0:nfs4_client as registered with the kstat 2992 * framework. 2993 */ 2994 static int 2995 cl4_snapshot(kstat_t *ksp, void *buf, int rw) 2996 { 2997 ksp->ks_snaptime = gethrtime(); 2998 if (rw == KSTAT_WRITE) { 2999 bcopy(buf, ksp->ks_private, sizeof (clstat4_tmpl)); 3000 } else { 3001 bcopy(ksp->ks_private, buf, sizeof (clstat4_tmpl)); 3002 } 3003 return (0); 3004 } 3005 3006 #ifdef DEBUG 3007 static int 3008 cl4_debug_snapshot(kstat_t *ksp, void *buf, int rw) 3009 { 3010 ksp->ks_snaptime = gethrtime(); 3011 if (rw == KSTAT_WRITE) { 3012 /* 3013 * Currently only the global zone can write to kstats, but we 3014 * add the check just for paranoia. 3015 */ 3016 if (INGLOBALZONE(curproc)) { 3017 bcopy(buf, &clstat4_debug, sizeof (clstat4_debug)); 3018 } 3019 } else { 3020 /* 3021 * If we're displaying the "global" debug kstat values, we 3022 * display them as-is to all zones since in fact they apply to 3023 * the system as a whole. 3024 */ 3025 bcopy(&clstat4_debug, buf, sizeof (clstat4_debug)); 3026 } 3027 return (0); 3028 } 3029 #endif 3030 3031 3032 3033 /* 3034 * Zone support 3035 */ 3036 static void * 3037 clinit4_zone(zoneid_t zoneid) 3038 { 3039 kstat_t *nfs4_client_kstat; 3040 kstat_t *nfs41_client_kstat; 3041 struct nfs4_clnt *nfscl; 3042 uint_t ndata; 3043 3044 nfscl = kmem_alloc(sizeof (*nfscl), KM_SLEEP); 3045 mutex_init(&nfscl->nfscl_chtable4_lock, NULL, MUTEX_DEFAULT, NULL); 3046 nfscl->nfscl_chtable4 = NULL; 3047 nfscl->nfscl_zoneid = zoneid; 3048 3049 bcopy(&clstat4_tmpl, &nfscl->nfscl_stat[NFS4_MINOR_v0], 3050 sizeof (clstat4_tmpl)); 3051 ndata = sizeof (clstat4_tmpl) / sizeof (kstat_named_t); 3052 if ((nfs4_client_kstat = kstat_create_zone("nfs", 0, "nfs4_client", 3053 "misc", KSTAT_TYPE_NAMED, ndata, 3054 KSTAT_FLAG_VIRTUAL | KSTAT_FLAG_WRITABLE, zoneid)) != NULL) { 3055 nfs4_client_kstat->ks_private = 3056 &nfscl->nfscl_stat[NFS4_MINOR_v0]; 3057 nfs4_client_kstat->ks_snapshot = cl4_snapshot; 3058 kstat_install(nfs4_client_kstat); 3059 } 3060 3061 bcopy(&clstat4_tmpl, &nfscl->nfscl_stat[NFS4_MINOR_v1], 3062 sizeof (clstat4_tmpl)); 3063 if ((nfs41_client_kstat = kstat_create_zone("nfs", 0, "nfs41_client", 3064 "misc", KSTAT_TYPE_NAMED, ndata, 3065 KSTAT_FLAG_VIRTUAL | KSTAT_FLAG_WRITABLE, zoneid)) != NULL) { 3066 nfs41_client_kstat->ks_private = 3067 &nfscl->nfscl_stat[NFS4_MINOR_v1]; 3068 nfs41_client_kstat->ks_snapshot = cl4_snapshot; 3069 kstat_install(nfs41_client_kstat); 3070 } 3071 3072 mutex_enter(&nfs4_clnt_list_lock); 3073 list_insert_head(&nfs4_clnt_list, nfscl); 3074 mutex_exit(&nfs4_clnt_list_lock); 3075 return (nfscl); 3076 } 3077 3078 /*ARGSUSED*/ 3079 static void 3080 clfini4_zone(zoneid_t zoneid, void *arg) 3081 { 3082 struct nfs4_clnt *nfscl = arg; 3083 chhead_t *chp, *next; 3084 3085 if (nfscl == NULL) 3086 return; 3087 mutex_enter(&nfs4_clnt_list_lock); 3088 list_remove(&nfs4_clnt_list, nfscl); 3089 mutex_exit(&nfs4_clnt_list_lock); 3090 clreclaim4_zone(nfscl, 0); 3091 for (chp = nfscl->nfscl_chtable4; chp != NULL; chp = next) { 3092 ASSERT(chp->ch_list == NULL); 3093 kmem_free(chp->ch_protofmly, strlen(chp->ch_protofmly) + 1); 3094 next = chp->ch_next; 3095 kmem_free(chp, sizeof (*chp)); 3096 } 3097 kstat_delete_byname_zone("nfs", 0, "nfs4_client", zoneid); 3098 kstat_delete_byname_zone("nfs", 0, "nfs41_client", zoneid); 3099 mutex_destroy(&nfscl->nfscl_chtable4_lock); 3100 kmem_free(nfscl, sizeof (*nfscl)); 3101 } 3102 3103 /* 3104 * Called by endpnt_destructor to make sure the client handles are 3105 * cleaned up before the RPC endpoints. This becomes a no-op if 3106 * clfini_zone (above) is called first. This function is needed 3107 * (rather than relying on clfini_zone to clean up) because the ZSD 3108 * callbacks have no ordering mechanism, so we have no way to ensure 3109 * that clfini_zone is called before endpnt_destructor. 3110 */ 3111 void 3112 clcleanup4_zone(zoneid_t zoneid) 3113 { 3114 struct nfs4_clnt *nfscl; 3115 3116 mutex_enter(&nfs4_clnt_list_lock); 3117 nfscl = list_head(&nfs4_clnt_list); 3118 for (; nfscl != NULL; nfscl = list_next(&nfs4_clnt_list, nfscl)) { 3119 if (nfscl->nfscl_zoneid == zoneid) { 3120 clreclaim4_zone(nfscl, 0); 3121 break; 3122 } 3123 } 3124 mutex_exit(&nfs4_clnt_list_lock); 3125 } 3126 3127 int 3128 nfs4_subr_init(void) 3129 { 3130 /* 3131 * Allocate and initialize the client handle cache 3132 */ 3133 #ifdef DEBUG 3134 uint_t ndata; 3135 kstat_t *nfs4_debug_kstat; 3136 #endif 3137 chtab4_cache = kmem_cache_create("client_handle4_cache", 3138 sizeof (struct chtab), 0, NULL, NULL, clreclaim4, NULL, 3139 NULL, 0); 3140 3141 #ifdef DEBUG 3142 /* 3143 * Create a kstat to maintain debug statistics across all zones 3144 */ 3145 ndata = sizeof (clstat4_debug) / sizeof (kstat_named_t); 3146 if ((nfs4_debug_kstat = kstat_create("nfs", 0, "nfs4_client_debug", 3147 "misc", KSTAT_TYPE_NAMED, ndata, 3148 KSTAT_FLAG_VIRTUAL | KSTAT_FLAG_WRITABLE)) != NULL) { 3149 nfs4_debug_kstat->ks_private = &clstat4_debug; 3150 nfs4_debug_kstat->ks_snapshot = cl4_debug_snapshot; 3151 kstat_install(nfs4_debug_kstat); 3152 } 3153 #endif 3154 3155 3156 /* 3157 * Initialize the list of per-zone client handles (and associated data). 3158 * This needs to be done before we call zone_key_create(). 3159 */ 3160 list_create(&nfs4_clnt_list, sizeof (struct nfs4_clnt), 3161 offsetof(struct nfs4_clnt, nfscl_node)); 3162 3163 /* 3164 * Initialize the zone_key for per-zone client handle lists. 3165 */ 3166 zone_key_create(&nfs4clnt_zone_key, clinit4_zone, NULL, clfini4_zone); 3167 3168 if (nfs4err_delay_time == 0) 3169 nfs4err_delay_time = NFS4ERR_DELAY_TIME; 3170 3171 return (0); 3172 } 3173 3174 int 3175 nfs4_subr_fini(void) 3176 { 3177 /* 3178 * Deallocate the client handle cache 3179 */ 3180 kmem_cache_destroy(chtab4_cache); 3181 #ifdef DEBUG 3182 kstat_delete_byname("nfs", 0, "nfs4_client_debug"); 3183 #endif 3184 3185 /* 3186 * Destroy the zone_key 3187 */ 3188 (void) zone_key_delete(nfs4clnt_zone_key); 3189 3190 return (0); 3191 } 3192 /* 3193 * Set or Clear direct I/O flag 3194 * VOP_RWLOCK() is held for write access to prevent a race condition 3195 * which would occur if a process is in the middle of a write when 3196 * directio flag gets set. It is possible that all pages may not get flushed. 3197 * 3198 * This is a copy of nfs_directio, changes here may need to be made 3199 * there and vice versa. 3200 */ 3201 3202 int 3203 nfs4_directio(vnode_t *vp, int cmd, cred_t *cr) 3204 { 3205 int error = 0; 3206 rnode4_t *rp; 3207 3208 rp = VTOR4(vp); 3209 3210 if (cmd == DIRECTIO_ON) { 3211 3212 if (rp->r_flags & R4DIRECTIO) 3213 return (0); 3214 3215 /* 3216 * Flush the page cache. 3217 */ 3218 3219 (void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, NULL); 3220 3221 if (rp->r_flags & R4DIRECTIO) { 3222 VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, NULL); 3223 return (0); 3224 } 3225 3226 if (nfs4_has_pages(vp) && 3227 ((rp->r_flags & R4DIRTY) || rp->r_awcount > 0)) { 3228 error = VOP_PUTPAGE(vp, (offset_t)0, (uint_t)0, 3229 B_INVAL, cr, NULL); 3230 if (error) { 3231 if (error == ENOSPC || error == EDQUOT) { 3232 mutex_enter(&rp->r_statelock); 3233 if (!rp->r_error) 3234 rp->r_error = error; 3235 mutex_exit(&rp->r_statelock); 3236 } 3237 VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, NULL); 3238 return (error); 3239 } 3240 } 3241 3242 mutex_enter(&rp->r_statelock); 3243 rp->r_flags |= R4DIRECTIO; 3244 mutex_exit(&rp->r_statelock); 3245 VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, NULL); 3246 return (0); 3247 } 3248 3249 if (cmd == DIRECTIO_OFF) { 3250 mutex_enter(&rp->r_statelock); 3251 rp->r_flags &= ~R4DIRECTIO; /* disable direct mode */ 3252 mutex_exit(&rp->r_statelock); 3253 return (0); 3254 } 3255 3256 return (EINVAL); 3257 } 3258 3259 /* 3260 * Return TRUE if the file has any pages. Always go back to 3261 * the master vnode to check v_pages since none of the shadows 3262 * can have pages. 3263 */ 3264 3265 bool_t 3266 nfs4_has_pages(vnode_t *vp) 3267 { 3268 rnode4_t *rp; 3269 3270 rp = VTOR4(vp); 3271 if (IS_SHADOW(vp, rp)) 3272 vp = RTOV4(rp); /* RTOV4 always gives the master */ 3273 3274 return (vn_has_cached_data(vp)); 3275 } 3276 3277 /* 3278 * This table is used to determine whether the client should attempt 3279 * failover based on the clnt_stat value returned by CLNT_CALL. The 3280 * clnt_stat is used as an index into the table. If 3281 * the error value that corresponds to the clnt_stat value in the 3282 * table is non-zero, then that is the error to be returned AND 3283 * that signals that failover should be attempted. 3284 * 3285 * Special note: If the RPC_ values change, then direct indexing of the 3286 * table is no longer valid, but having the RPC_ values in the table 3287 * allow the functions to detect the change and issue a warning. 3288 * In this case, the code will always attempt failover as a defensive 3289 * measure. 3290 */ 3291 3292 static struct try_failover_tab { 3293 enum clnt_stat cstat; 3294 int error; 3295 } try_failover_table [] = { 3296 3297 RPC_SUCCESS, 0, 3298 RPC_CANTENCODEARGS, 0, 3299 RPC_CANTDECODERES, 0, 3300 RPC_CANTSEND, ECOMM, 3301 RPC_CANTRECV, ECOMM, 3302 RPC_TIMEDOUT, ETIMEDOUT, 3303 RPC_VERSMISMATCH, 0, 3304 RPC_AUTHERROR, 0, 3305 RPC_PROGUNAVAIL, 0, 3306 RPC_PROGVERSMISMATCH, 0, 3307 RPC_PROCUNAVAIL, 0, 3308 RPC_CANTDECODEARGS, 0, 3309 RPC_SYSTEMERROR, ENOSR, 3310 RPC_UNKNOWNHOST, EHOSTUNREACH, 3311 RPC_RPCBFAILURE, ENETUNREACH, 3312 RPC_PROGNOTREGISTERED, ECONNREFUSED, 3313 RPC_FAILED, ETIMEDOUT, 3314 RPC_UNKNOWNPROTO, EHOSTUNREACH, 3315 RPC_INTR, 0, 3316 RPC_UNKNOWNADDR, EHOSTUNREACH, 3317 RPC_TLIERROR, 0, 3318 RPC_NOBROADCAST, EHOSTUNREACH, 3319 RPC_N2AXLATEFAILURE, ECONNREFUSED, 3320 RPC_UDERROR, 0, 3321 RPC_INPROGRESS, 0, 3322 RPC_STALERACHANDLE, EINVAL, 3323 RPC_CANTCONNECT, ECONNREFUSED, 3324 RPC_XPRTFAILED, ECONNABORTED, 3325 RPC_CANTCREATESTREAM, ECONNREFUSED, 3326 RPC_CANTSTORE, ENOBUFS, 3327 RPC_CONN_NOT_BOUND, 0 3328 }; 3329 3330 /* 3331 * nfs4_try_failover - determine whether the client should 3332 * attempt failover based on the values stored in the nfs4_error_t. 3333 */ 3334 int 3335 nfs4_try_failover(nfs4_error_t *ep) 3336 { 3337 if (ep->error == ETIMEDOUT || ep->stat == NFS4ERR_RESOURCE) 3338 return (TRUE); 3339 3340 if (ep->error && ep->rpc_status != RPC_SUCCESS) 3341 return (try_failover(ep->rpc_status) != 0 ? TRUE : FALSE); 3342 3343 return (FALSE); 3344 } 3345 3346 /* 3347 * try_failover - internal version of nfs4_try_failover, called 3348 * only by rfscall and aclcall. Determine if failover is warranted 3349 * based on the clnt_stat and return the error number if it is. 3350 */ 3351 static int 3352 try_failover(enum clnt_stat rpc_status) 3353 { 3354 int err = 0; 3355 3356 if (rpc_status == RPC_SUCCESS) 3357 return (0); 3358 3359 #ifdef DEBUG 3360 if (rpc_status != 0 && nfs4_try_failover_any) { 3361 err = ETIMEDOUT; 3362 goto done; 3363 } 3364 #endif 3365 /* 3366 * The rpc status is used as an index into the table. 3367 * If the rpc status is outside of the range of the 3368 * table or if the rpc error numbers have been changed 3369 * since the table was constructed, then print a warning 3370 * (DEBUG only) and try failover anyway. Otherwise, just 3371 * grab the resulting error number out of the table. 3372 */ 3373 if (rpc_status < RPC_SUCCESS || rpc_status >= 3374 sizeof (try_failover_table)/sizeof (try_failover_table[0]) || 3375 try_failover_table[rpc_status].cstat != rpc_status) { 3376 3377 err = ETIMEDOUT; 3378 #ifdef DEBUG 3379 cmn_err(CE_NOTE, "try_failover: unexpected rpc error %d", 3380 rpc_status); 3381 #endif 3382 } else 3383 err = try_failover_table[rpc_status].error; 3384 3385 done: 3386 if (rpc_status) 3387 NFS4_DEBUG(nfs4_client_failover_debug, (CE_NOTE, 3388 "nfs4_try_failover: %strying failover on error %d", 3389 err ? "" : "NOT ", rpc_status)); 3390 3391 return (err); 3392 } 3393 3394 void 3395 nfs4_error_zinit(nfs4_error_t *ep) 3396 { 3397 ep->error = 0; 3398 ep->stat = NFS4_OK; 3399 ep->rpc_status = RPC_SUCCESS; 3400 } 3401 3402 void 3403 nfs4_error_init(nfs4_error_t *ep, int error) 3404 { 3405 ep->error = error; 3406 ep->stat = NFS4_OK; 3407 ep->rpc_status = RPC_SUCCESS; 3408 } 3409 3410 3411 #ifdef DEBUG 3412 3413 /* 3414 * Return a 16-bit hash for filehandle, stateid, clientid, owner. 3415 * use the same algorithm as for NFS v3. 3416 * 3417 */ 3418 int 3419 hash16(void *p, int len) 3420 { 3421 int i, rem; 3422 uint_t *wp; 3423 uint_t key = 0; 3424 3425 /* protect against non word aligned */ 3426 if ((rem = len & 3) != 0) 3427 len &= ~3; 3428 3429 for (i = 0, wp = (uint_t *)p; i < len; i += 4, wp++) { 3430 key ^= (*wp >> 16) ^ *wp; 3431 } 3432 3433 /* hash left-over bytes */ 3434 for (i = 0; i < rem; i++) 3435 key ^= *((uchar_t *)p + i); 3436 3437 return (key & 0xffff); 3438 } 3439 3440 /* 3441 * rnode4info - return filehandle and path information for an rnode. 3442 * XXX MT issues: uses a single static buffer, no locking of path. 3443 */ 3444 char * 3445 rnode4info(rnode4_t *rp) 3446 { 3447 static char buf[80]; 3448 nfs4_fhandle_t fhandle; 3449 char *path; 3450 char *type; 3451 3452 if (rp == NULL) 3453 return ("null"); 3454 if (rp->r_flags & R4ISXATTR) 3455 type = "attr"; 3456 else if (RTOV4(rp)->v_flag & V_XATTRDIR) 3457 type = "attrdir"; 3458 else if (RTOV4(rp)->v_flag & VROOT) 3459 type = "root"; 3460 else if (RTOV4(rp)->v_type == VDIR) 3461 type = "dir"; 3462 else if (RTOV4(rp)->v_type == VREG) 3463 type = "file"; 3464 else 3465 type = "other"; 3466 sfh4_copyval(rp->r_fh, &fhandle); 3467 path = fn_path(rp->r_svnode.sv_name); 3468 (void) snprintf(buf, 80, "$%p[%s], type=%s, flags=%04X, FH=%04X\n", 3469 (void *)rp, path, type, rp->r_flags, 3470 hash16((void *)&fhandle.fh_buf, fhandle.fh_len)); 3471 kmem_free(path, strlen(path)+1); 3472 return (buf); 3473 } 3474 #endif 3475 3476 int nfs4_sessions_debug; 3477 3478 void 3479 nfs4sequence_setup(nfs4_session_t *np, COMPOUND4args_clnt *rfsargp, 3480 nfs4_slot_t **slotpp) 3481 { 3482 int slot_id = 0; 3483 nfs4_slot_t *slot; 3484 3485 bcopy(&np->sessionid, 3486 rfsargp->array->nfs_argop4_u.opsequence.sa_sessionid, 3487 sizeof (sessionid4)); 3488 3489 /* 3490 * Find a slot to use. 3491 */ 3492 (void) nfs_rw_enter_sig(&np->slot_table_rwlock, RW_READER, 0); 3493 mutex_enter(&np->slot_lock); 3494 slot_id = np->next_slot; 3495 while ((np->slot_table[slot_id]->slot_inuse != 0) || 3496 (np->slot_table[slot_id]->slot_bad != 0)) { 3497 /* 3498 * Can drop the rwlock here so we don't hold it over 3499 * a possible cv_wait. 3500 */ 3501 nfs_rw_exit(&np->slot_table_rwlock); 3502 3503 /* 3504 * This slot is still in use. 3505 * Check next slot if there are still some available. 3506 */ 3507 3508 while (np->slots_available == 0) { 3509 if (nfs4_sessions_debug) 3510 cmn_err(CE_WARN, "Waiting for Available Slot"); 3511 cv_wait(&np->slot_wait, &np->slot_lock); 3512 } 3513 slot_id++; 3514 if (slot_id == np->maxslots) 3515 slot_id = 0; 3516 (void) nfs_rw_enter_sig(&np->slot_table_rwlock, RW_READER, 0); 3517 } 3518 *slotpp = slot = np->slot_table[slot_id]; 3519 slot->slot_inuse = 1; 3520 np->slots_available--; 3521 np->next_slot = slot_id + 1 == np->maxslots ? 0 : slot_id + 1; 3522 3523 /* 3524 * Update SEQUENCE args 3525 */ 3526 rfsargp->array->nfs_argop4_u.opsequence.sa_sequenceid = 3527 slot->slot_seqid; 3528 rfsargp->array->nfs_argop4_u.opsequence.sa_slotid = slot->slot_id; 3529 rfsargp->array->nfs_argop4_u.opsequence.sa_highest_slotid = 3530 np->maxslots - np->slots_available; 3531 /* XXX - rick - need sr_target_highest_slotid */ 3532 mutex_exit(&np->slot_lock); 3533 nfs_rw_exit(&np->slot_table_rwlock); 3534 } 3535 3536 void 3537 nfs4sequence_fin(nfs4_session_t *np, COMPOUND4res_clnt *rfsresp, 3538 nfs4_slot_t *slot, nfs4_error_t *ep) 3539 { 3540 SEQUENCE4resok *seqres; 3541 3542 mutex_enter(&np->slot_lock); 3543 3544 ASSERT(slot->slot_inuse); 3545 slot->slot_inuse = 0; 3546 3547 /* if call started but not completed, mark slot as bad */ 3548 if ((ep->error != 0) && 3549 ((ep->rpc_status == RPC_TIMEDOUT) || 3550 (ep->rpc_status == RPC_INTR))) { 3551 cmn_err(CE_WARN, "SEQUENCE failed %d, bad slot %d:%d", 3552 ep->rpc_status, slot->slot_id, slot->slot_seqid); 3553 slot->slot_bad = 1; 3554 } else { 3555 if (slot->slot_id < np->next_slot) 3556 np->next_slot = slot->slot_id; 3557 3558 /* Update slot seqid on successful op_sequence */ 3559 if (ep->error == 0 && (rfsresp->array != NULL && 3560 rfsresp->array->nfs_resop4_u.opsequence.sr_status == 3561 NFS4_OK)) 3562 slot->slot_seqid++; 3563 3564 if (np->slots_available++ == 0) { 3565 if (nfs4_sessions_debug) 3566 cmn_err(CE_WARN, "Slots Available"); 3567 cv_broadcast(&np->slot_wait); 3568 } 3569 } 3570 3571 mutex_exit(&np->slot_lock); 3572 3573 /* SEQUENCE Op Successful? */ 3574 if (ep->error != 0 || rfsresp->status != NFS4_OK || 3575 (rfsresp->array != NULL && 3576 rfsresp->array->nfs_resop4_u.opsequence.sr_status != NFS4_OK)) { 3577 /* 3578 * cmn_err(CE_WARN, "sequence op failed or missing\n"); 3579 */ 3580 return; 3581 } 3582 3583 seqres = &rfsresp->array->nfs_resop4_u.opsequence. 3584 SEQUENCE4res_u.sr_resok4; 3585 3586 /* 3587 * Sequence Op Successful, Handle Errors and maxslot changes. 3588 */ 3589 3590 if (seqres->sr_status_flags & SEQ4_STATUS_CB_PATH_DOWN) { 3591 #ifdef notyet 3592 nfs4_delegreturn_all(np); 3593 #else 3594 cmn_err(CE_WARN, "SEQ4_STATUS_CB_PATH_DOWN not handled"); 3595 #endif 3596 } 3597 3598 if (seqres->sr_status_flags & SEQ4_STATUS_CB_GSS_CONTEXTS_EXPIRING) { 3599 cmn_err(CE_WARN, "SEQUENCE got CB_GSS_CONTEXTS_EXPIRING"); 3600 } 3601 3602 if (seqres->sr_status_flags & SEQ4_STATUS_CB_GSS_CONTEXTS_EXPIRED) { 3603 cmn_err(CE_WARN, "SEQUENCE got CB_GSS_CONTEXTS_EXPIRED"); 3604 } 3605 3606 if (seqres->sr_status_flags & SEQ4_STATUS_EXPIRED_ALL_STATE_REVOKED) { 3607 cmn_err(CE_WARN, "SEQUENCE got EXIPRED_ALL_STATE_REVOKED"); 3608 } 3609 3610 if (seqres->sr_status_flags & SEQ4_STATUS_EXPIRED_SOME_STATE_REVOKED) { 3611 cmn_err(CE_WARN, "SEQUENCE got EXPIRED_SOME_STATE_REVOKED"); 3612 } 3613 3614 if (seqres->sr_status_flags & SEQ4_STATUS_ADMIN_STATE_REVOKED) { 3615 cmn_err(CE_WARN, "SEQUENCE got ADMIN_STATE_REVOKED"); 3616 } 3617 3618 if (seqres->sr_status_flags & SEQ4_STATUS_RECALLABLE_STATE_REVOKED) { 3619 cmn_err(CE_WARN, "SEQUENCE got RECALLABLE_STATE_REVOKED"); 3620 } 3621 3622 if (seqres->sr_status_flags & SEQ4_STATUS_LEASE_MOVED) { 3623 cmn_err(CE_WARN, "SEQUENCE got LEASE_MOVED"); 3624 } 3625 } 3626 3627 kmutex_t nfs4_session_lst_lock; 3628 list_t nfs4_session_list; 3629 3630 void 3631 nfs4session_init() 3632 { 3633 mutex_init(&nfs4_session_lst_lock, NULL, MUTEX_DEFAULT, NULL); 3634 list_create(&nfs4_session_list, sizeof (nfs4_session_t), 3635 offsetof(nfs4_session_t, ssx_list)); 3636 } 3637 3638 /* 3639 * Compare 2 netbufs, return true of they match 3640 */ 3641 int 3642 netbuf_match(struct netbuf *n1, struct netbuf *n2) 3643 { 3644 if (n1->len == n2->len && bcmp(n1->buf, n2->buf, n1->len) == 0) 3645 return (1); 3646 return (0); 3647 } 3648 3649 void * 3650 new_string(void *cur) 3651 { 3652 void *v; 3653 3654 v = kmem_alloc(strlen(cur)+1, KM_SLEEP); 3655 (void) strcpy(v, cur); 3656 return (v); 3657 } 3658 3659 servinfo4_t * 3660 new_servinfo4(struct knetconfig *knc, struct netbuf *nb, int flags) 3661 { 3662 servinfo4_t *svp; 3663 struct sec_data *secdata; 3664 3665 /* 3666 * Allocate a servinfo4 struct. 3667 */ 3668 svp = kmem_zalloc(sizeof (*svp), KM_SLEEP); 3669 nfs_rw_init(&svp->sv_lock, NULL, RW_DEFAULT, NULL); 3670 svp->sv_flags = flags; 3671 3672 svp->sv_knconf = kmem_alloc(sizeof (*knc), KM_SLEEP); 3673 svp->sv_knconf->knc_semantics = knc->knc_semantics; 3674 svp->sv_knconf->knc_protofmly = new_string(knc->knc_protofmly); 3675 svp->sv_knconf->knc_proto = new_string(knc->knc_proto); 3676 svp->sv_knconf->knc_rdev = knc->knc_rdev; 3677 bzero(svp->sv_knconf->knc_unused, sizeof (knc->knc_unused)); 3678 3679 svp->sv_addr.maxlen = nb->maxlen; 3680 svp->sv_addr.len = nb->len; 3681 svp->sv_addr.buf = kmem_alloc(nb->maxlen, KM_SLEEP); 3682 bcopy(nb->buf, svp->sv_addr.buf, nb->len); 3683 3684 /* XXX, ought to inherit sec data from parent servinfo4 */ 3685 secdata = kmem_alloc(sizeof (*secdata), KM_SLEEP); 3686 secdata->secmod = secdata->rpcflavor = AUTH_SYS; 3687 secdata->data = NULL; 3688 svp->sv_secdata = secdata; 3689 3690 /* XXX */ 3691 svp->sv_path = "/"; 3692 svp->sv_pathlen = 1; 3693 svp->sv_hostname = "data-server"; 3694 svp->sv_hostnamelen = strlen("data-server"); 3695 3696 return (svp); 3697 } 3698 3699 /* 3700 * XXX - this will be eliminated once everyone is calling rfs4call() 3701 * emulate the behavior of rfs4call for those who call 3702 * CLNT_CALL directly 3703 */ 3704 void 3705 nfs4_error_set(nfs4_error_t *ep, enum clnt_stat rpc_status, enum nfsstat4 stat) 3706 { 3707 if (rpc_status == RPC_SUCCESS) { 3708 ep->error = 0; /* geterrno4 happens higher up */ 3709 ep->stat = stat; 3710 ep->rpc_status = RPC_SUCCESS; 3711 } else { 3712 ep->error = EPROTO; /* XXX */ 3713 ep->stat = 0; 3714 ep->rpc_status = rpc_status; 3715 } 3716 } 3717 3718 /* 3719 * A function to interface with RPC tags. 3720 * Returns 0 on success 3721 */ 3722 int 3723 nfs4_tag_ctl(nfs4_server_t *np, mntinfo4_t *mi, servinfo4_t *svp, 3724 sessionid4 oldsid, int cmd, cred_t *cr) 3725 { 3726 int error; 3727 CLIENT *client; 3728 struct chtab *ch; 3729 struct nfs4_clnt *nfscl; 3730 3731 nfscl = zone_getspecific(nfs4clnt_zone_key, nfs_zone()); 3732 ASSERT(nfscl != NULL); 3733 3734 if (svp == NULL) { 3735 /* 3736 * We just pick the current servinfo ptr. Even if 3737 * this changes midstream, we should be alright, since 3738 * we are not really going OTW. Just used to get a 3739 * client handle. 3740 */ 3741 mutex_enter(&mi->mi_lock); 3742 svp = mi->mi_curr_serv; 3743 mutex_exit(&mi->mi_lock); 3744 } 3745 3746 error = nfs_clget4(mi, svp, cr, &client, &ch, nfscl); 3747 3748 if (error) 3749 return (error); 3750 3751 switch (cmd) { 3752 case NFS4_TAG_SWAP: 3753 3754 /* 3755 * To do the sessid swap first set the old tag and 3756 * then call to swap to the new one 3757 */ 3758 3759 if (!CLNT_CONTROL(client, CLSET_TAG, (char *)oldsid)) { 3760 zcmn_err(getzoneid(), CE_WARN, 3761 "Failed to set tag on client handle"); 3762 error = EIO; 3763 break; 3764 } 3765 3766 /* 3767 * This switches the tag value in the RPC layer 3768 * The client handle's tag (client->cku_tag) is set 3769 * to new tag as well. 3770 */ 3771 3772 if (!CLNT_CONTROL(client, CLSET_TAG_SWAP, 3773 (char *)(np->ssx.sessionid))) { 3774 zcmn_err(getzoneid(), CE_WARN, 3775 "Failed to swap rpc tags"); 3776 error = EIO; 3777 } 3778 3779 break; 3780 3781 case NFS4_TAG_DESTROY: 3782 3783 if (!CLNT_CONTROL(client, CLSET_TAG_DESTROY, 3784 (char *)(np->ssx.sessionid))) { 3785 zcmn_err(getzoneid(), CE_WARN, 3786 "Failed destroy rpc tags"); 3787 error = EIO; 3788 } 3789 break; 3790 3791 case NFS4_CBSERVER_CLEANUP: 3792 if (!CLNT_CONTROL(client, CLSET_CBSERVER_CLEANUP, 3793 (char *)(np->ssx.sessionid))) { 3794 zcmn_err(getzoneid(), CE_WARN, 3795 "Failed destroy rpc tags"); 3796 error = EIO; 3797 } 3798 break; 3799 } 3800 3801 clfree4(client, ch, nfscl); 3802 return (error); 3803 } 3804 3805 /* 3806 * All NFSv4.1 defined errors 3807 */ 3808 char * 3809 nfs41_strerror(nfsstat4 err) 3810 { 3811 switch (err) { 3812 case NFS4_OK: 3813 return ("NFS4_OK"); 3814 case NFS4ERR_PERM: 3815 return ("NFS4ERR_PERM"); 3816 case NFS4ERR_NOENT: 3817 return ("NFS4ERR_NOENT"); 3818 case NFS4ERR_IO: 3819 return ("NFS4ERR_IO"); 3820 case NFS4ERR_NXIO: 3821 return ("NFS4ERR_NXIO"); 3822 case NFS4ERR_ACCESS: 3823 return ("NFS4ERR_ACCESS"); 3824 case NFS4ERR_EXIST: 3825 return ("NFS4ERR_EXIST"); 3826 case NFS4ERR_XDEV: 3827 return ("NFS4ERR_XDEV"); 3828 case NFS4ERR_NOTDIR: 3829 return ("NFS4ERR_NOTDIR"); 3830 case NFS4ERR_ISDIR: 3831 return ("NFS4ERR_ISDIR"); 3832 case NFS4ERR_INVAL: 3833 return ("NFS4ERR_INVAL"); 3834 case NFS4ERR_FBIG: 3835 return ("NFS4ERR_FBIG"); 3836 case NFS4ERR_NOSPC: 3837 return ("NFS4ERR_NOSPC"); 3838 case NFS4ERR_ROFS: 3839 return ("NFS4ERR_ROFS"); 3840 case NFS4ERR_MLINK: 3841 return ("NFS4ERR_MLINK"); 3842 case NFS4ERR_NAMETOOLONG: 3843 return ("NFS4ERR_NAMETOOLONG"); 3844 case NFS4ERR_NOTEMPTY: 3845 return ("NFS4ERR_NOTEMPTY"); 3846 case NFS4ERR_DQUOT: 3847 return ("NFS4ERR_DQUOT"); 3848 case NFS4ERR_STALE: 3849 return ("NFS4ERR_STALE"); 3850 case NFS4ERR_BADHANDLE: 3851 return ("NFS4ERR_BADHANDLE"); 3852 case NFS4ERR_BAD_COOKIE: 3853 return ("NFS4ERR_BAD_COOKIE"); 3854 case NFS4ERR_NOTSUPP: 3855 return ("NFS4ERR_NOTSUPP"); 3856 case NFS4ERR_TOOSMALL: 3857 return ("NFS4ERR_TOOSMALL"); 3858 case NFS4ERR_SERVERFAULT: 3859 return ("NFS4ERR_SERVERFAULT"); 3860 case NFS4ERR_BADTYPE: 3861 return ("NFS4ERR_BADTYPE"); 3862 case NFS4ERR_DELAY: 3863 return ("NFS4ERR_DELAY"); 3864 case NFS4ERR_SAME: 3865 return ("NFS4ERR_SAME"); 3866 case NFS4ERR_DENIED: 3867 return ("NFS4ERR_DENIED"); 3868 case NFS4ERR_EXPIRED: 3869 return ("NFS4ERR_EXPIRED"); 3870 case NFS4ERR_LOCKED: 3871 return ("NFS4ERR_LOCKED"); 3872 case NFS4ERR_GRACE: 3873 return ("NFS4ERR_GRACE"); 3874 case NFS4ERR_FHEXPIRED: 3875 return ("NFS4ERR_FHEXPIRED"); 3876 case NFS4ERR_SHARE_DENIED: 3877 return ("NFS4ERR_SHARE_DENIED"); 3878 case NFS4ERR_WRONGSEC: 3879 return ("NFS4ERR_WRONGSEC"); 3880 case NFS4ERR_CLID_INUSE: 3881 return ("NFS4ERR_CLID_INUSE"); 3882 case NFS4ERR_RESOURCE: 3883 return ("NFS4ERR_RESOURCE"); 3884 case NFS4ERR_MOVED: 3885 return ("NFS4ERR_MOVED"); 3886 case NFS4ERR_NOFILEHANDLE: 3887 return ("NFS4ERR_NOFILEHANDLE"); 3888 case NFS4ERR_MINOR_VERS_MISMATCH: 3889 return ("NFS4ERR_MINOR_VERS_MISMATCH"); 3890 case NFS4ERR_STALE_CLIENTID: 3891 return ("NFS4ERR_STALE_CLIENTID"); 3892 case NFS4ERR_STALE_STATEID: 3893 return ("NFS4ERR_STALE_STATEID"); 3894 case NFS4ERR_OLD_STATEID: 3895 return ("NFS4ERR_OLD_STATEID"); 3896 case NFS4ERR_BAD_STATEID: 3897 return ("NFS4ERR_BAD_STATEID"); 3898 case NFS4ERR_BAD_SEQID: 3899 return ("NFS4ERR_BAD_SEQID"); 3900 case NFS4ERR_NOT_SAME: 3901 return ("NFS4ERR_NOT_SAME"); 3902 case NFS4ERR_LOCK_RANGE: 3903 return ("NFS4ERR_LOCK_RANGE"); 3904 case NFS4ERR_SYMLINK: 3905 return ("NFS4ERR_SYMLINK"); 3906 case NFS4ERR_RESTOREFH: 3907 return ("NFS4ERR_RESTOREFH"); 3908 case NFS4ERR_LEASE_MOVED: 3909 return ("NFS4ERR_LEASE_MOVED"); 3910 case NFS4ERR_ATTRNOTSUPP: 3911 return ("NFS4ERR_ATTRNOTSUPP"); 3912 case NFS4ERR_NO_GRACE: 3913 return ("NFS4ERR_NO_GRACE"); 3914 case NFS4ERR_RECLAIM_BAD: 3915 return ("NFS4ERR_RECLAIM_BAD"); 3916 case NFS4ERR_RECLAIM_CONFLICT: 3917 return ("NFS4ERR_RECLAIM_CONFLICT"); 3918 case NFS4ERR_BADXDR: 3919 return ("NFS4ERR_BADXDR"); 3920 case NFS4ERR_LOCKS_HELD: 3921 return ("NFS4ERR_LOCKS_HELD"); 3922 case NFS4ERR_OPENMODE: 3923 return ("NFS4ERR_OPENMODE"); 3924 case NFS4ERR_BADOWNER: 3925 return ("NFS4ERR_BADOWNER"); 3926 case NFS4ERR_BADCHAR: 3927 return ("NFS4ERR_BADCHAR"); 3928 case NFS4ERR_BADNAME: 3929 return ("NFS4ERR_BADNAME"); 3930 case NFS4ERR_BAD_RANGE: 3931 return ("NFS4ERR_BAD_RANGE"); 3932 case NFS4ERR_LOCK_NOTSUPP: 3933 return ("NFS4ERR_LOCK_NOTSUPP"); 3934 case NFS4ERR_OP_ILLEGAL: 3935 return ("NFS4ERR_OP_ILLEGAL"); 3936 case NFS4ERR_DEADLOCK: 3937 return ("NFS4ERR_DEADLOCK"); 3938 case NFS4ERR_FILE_OPEN: 3939 return ("NFS4ERR_FILE_OPEN"); 3940 case NFS4ERR_ADMIN_REVOKED: 3941 return ("NFS4ERR_ADMIN_REVOKED"); 3942 case NFS4ERR_CB_PATH_DOWN: 3943 return ("NFS4ERR_CB_PATH_DOWN"); 3944 case NFS4ERR_BADIOMODE: 3945 return ("NFS4ERR_BADIOMODE"); 3946 case NFS4ERR_BADLAYOUT: 3947 return ("NFS4ERR_BADLAYOUT"); 3948 case NFS4ERR_BAD_SESSION_DIGEST: 3949 return ("NFS4ERR_BAD_SESSION_DIGEST"); 3950 case NFS4ERR_BADSESSION: 3951 return ("NFS4ERR_BADSESSION"); 3952 case NFS4ERR_BADSLOT: 3953 return ("NFS4ERR_BADSLOT"); 3954 case NFS4ERR_COMPLETE_ALREADY: 3955 return ("NFS4ERR_COMPLETE_ALREADY"); 3956 case NFS4ERR_CONN_NOT_BOUND_TO_SESSION: 3957 return ("NFS4ERR_CONN_NOT_BOUND_TO_SESSION"); 3958 case NFS4ERR_DELEG_ALREADY_WANTED: 3959 return ("NFS4ERR_DELEG_ALREADY_WANTED"); 3960 case NFS4ERR_BACK_CHAN_BUSY: 3961 return ("NFS4ERR_BACK_CHAN_BUSY"); 3962 case NFS4ERR_LAYOUTTRYLATER: 3963 return ("NFS4ERR_LAYOUTTRYLATER"); 3964 case NFS4ERR_LAYOUTUNAVAILABLE: 3965 return ("NFS4ERR_LAYOUTUNAVAILABLE"); 3966 case NFS4ERR_NOMATCHING_LAYOUT: 3967 return ("NFS4ERR_NOMATCHING_LAYOUT"); 3968 case NFS4ERR_RECALLCONFLICT: 3969 return ("NFS4ERR_RECALLCONFLICT"); 3970 case NFS4ERR_UNKNOWN_LAYOUTTYPE: 3971 return ("NFS4ERR_UNKNOWN_LAYOUTTYPE"); 3972 case NFS4ERR_SEQ_MISORDERED: 3973 return ("NFS4ERR_SEQ_MISORDERED"); 3974 case NFS4ERR_SEQUENCE_POS: 3975 return ("NFS4ERR_SEQUENCE_POS"); 3976 case NFS4ERR_REQ_TOO_BIG: 3977 return ("NFS4ERR_REQ_TOO_BIG"); 3978 case NFS4ERR_REP_TOO_BIG: 3979 return ("NFS4ERR_REP_TOO_BIG"); 3980 case NFS4ERR_REP_TOO_BIG_TO_CACHE: 3981 return ("NFS4ERR_REP_TOO_BIG_TO_CACHE"); 3982 case NFS4ERR_RETRY_UNCACHED_REP: 3983 return ("NFS4ERR_RETRY_UNCACHED_REP"); 3984 case NFS4ERR_UNSAFE_COMPOUND: 3985 return ("NFS4ERR_UNSAFE_COMPOUND"); 3986 case NFS4ERR_TOO_MANY_OPS: 3987 return ("NFS4ERR_TOO_MANY_OPS"); 3988 case NFS4ERR_OP_NOT_IN_SESSION: 3989 return ("NFS4ERR_OP_NOT_IN_SESSION"); 3990 case NFS4ERR_HASH_ALG_UNSUPP: 3991 return ("NFS4ERR_HASH_ALG_UNSUPP"); 3992 case NFS4ERR_CLIENTID_BUSY: 3993 return ("NFS4ERR_CLIENTID_BUSY"); 3994 case NFS4ERR_PNFS_IO_HOLE: 3995 return ("NFS4ERR_PNFS_IO_HOLE"); 3996 case NFS4ERR_SEQ_FALSE_RETRY: 3997 return ("NFS4ERR_SEQ_FALSE_RETRY"); 3998 case NFS4ERR_BAD_HIGH_SLOT: 3999 return ("NFS4ERR_BAD_HIGH_SLOT"); 4000 case NFS4ERR_DEADSESSION: 4001 return ("NFS4ERR_DEADSESSION"); 4002 case NFS4ERR_ENCR_ALG_UNSUPP: 4003 return ("NFS4ERR_ENCR_ALG_UNSUPP"); 4004 case NFS4ERR_PNFS_NO_LAYOUT: 4005 return ("NFS4ERR_PNFS_NO_LAYOUT"); 4006 case NFS4ERR_NOT_ONLY_OP: 4007 return ("NFS4ERR_NOT_ONLY_OP"); 4008 case NFS4ERR_WRONG_CRED: 4009 return ("NFS4ERR_WRONG_CRED"); 4010 case NFS4ERR_WRONG_TYPE: 4011 return ("NFS4ERR_WRONG_TYPE"); 4012 case NFS4ERR_DIRDELEG_UNAVAIL: 4013 return ("NFS4ERR_DIRDELEG_UNAVAIL"); 4014 case NFS4ERR_REJECT_DELEG: 4015 return ("NFS4ERR_REJECT_DELEG"); 4016 case NFS4ERR_RETURNCONFLICT: 4017 return ("NFS4ERR_RETURNCONFLICT"); 4018 default: 4019 { 4020 static char msg[99]; 4021 static char *ies = "Unknown NFSv4.1 error"; 4022 4023 (void) snprintf(msg, 99, "%s: %d", ies, (int)err); 4024 return (msg); 4025 } 4026 } 4027 } --- EOF ---